Research ArticleMobility Similarity-Based Routing in Buffer-LimitedDelay Tolerant Networks

Yao Liu123 Jiawei Huang1 Weiping Wang1 Hongjing Zhou2 Ying An1 and Jianxin Wang1

1School of Information Science and Engineering Central South University Changsha 410083 China2School of Computer and Information Engineering Hunan University of Commerce Changsha 410205 China3Science and Technology on Information Transmission and Dissemination in Communication Networks LaboratoryShijiazhuang Hebei 050081 China

Correspondence should be addressed to Jiawei Huang jiaweihuangcsueducn

Received 19 December 2014 Revised 13 April 2015 Accepted 17 April 2015

Academic Editor Alessandro Nordio

Copyright copy 2015 Yao Liu et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

In delay tolerant networks (DTNs) multiple-copy routing is often adopted to improve the probability of successful messagedelivery but causes more traffic loadsThe excessive increase in multiple messages copies often exhausts the network resources anddeteriorates its performance significantly To solve this problem a mobility similarity-based routing (SBR) algorithm is proposedin this paper On one hand destination similarity is used to help the message carrier node find the appropriate relay nodes withthe higher opportunity to meet the destination node in order to improve the delivery ratio On the other hand carrier similarityis adopted to avoid sending the same copy to the nodes with mobility similarity so as to reduce the number of message copiesFurthermore inspired by the law of diminishing marginal utility in economics a buffer management scheme based on the messagetransmission status is proposed Experimental results show that the proposed SBR routing algorithm combined with the buffermanagement scheme can improve the delivery ratio and has the lower overhead ratio compared to other routing algorithms

1 Introduction

Delay tolerant networks (DTNs) [1] are mobile wirelessnetworks that experience frequent partitions In DTNs theremay never exist a fully connected path from a source toa destination The basic assumption on stable end-to-endpath of routing protocols is invalidMany real-world networkscenarios fall into this paradigm such as vehicular ad hocnetworks [2] sensor networks for wildlife tracking andhabitat monitoring [3] pocket switched networks [4] andmobile social networks [5]

In DTNs communication is mainly dependent on thenode mobility which plays an important role in overcom-ing the lack of end-to-end path The storage-carry-forward(SCF) routing paradigm is used to deliver messages in thechallenging environments Specifically if a node receives amessage from an encountered node it stores and carries themessage until another communication opportunity arisesDepending on the method whether the message carrier node

keeps or removes the forwarded message there are twoclasses of routing scheme in DTNs single-copy routing [6]and multiple-copy routing [7]

As the primary purpose of DTN routing is to select someproper relaying nodes to guarantee that the message willbe forwarded to the destination the DTN routing protocolsusually use themultiple-copy scheme to provide high deliveryratio But the multiple-copy schemes cause tremendousnetwork overheads How to achieve high delivery ratioand maintain low overheads at the same time is the mostimportant research issue for multiple-copy routing

Moreover the routing schemes are usually designed onlyfrom the viewpoint of single message and do not considerthe problem of the usage of the whole network resourcesMany proposed routing schemes forDTNs assume that nodeshave infinite buffer space and ignore the contention forbuffer space in network nodes In practice the nodes havelimited buffer space in many wireless networks Even if anode has large buffer space it may share only a limited small

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2015 Article ID 593607 16 pageshttpdxdoiorg1011552015593607

2 International Journal of Distributed Sensor Networks

part with external traffic when acting as a relaying nodeThus how to manage buffer space significantly affects theperformance of multiple-copy routing protocols especiallyin the environments where intermittent connectivity andlong latency require the data to be stored for a long periodthroughout the network

In this paper we firstly propose a similarity-based routingalgorithm in DTNs Two mobility similarity metrics areadopted that is destination similarity and carrier similarityto choose the next relay nodeThe destination similarity refersto the mobility similarity between an intermediate node andthe destination The carrier similarity refers to the mobilitysimilarity between the message carrier node and its encoun-tered node In addition to using the destination similarity toimprove message delivery probability the carrier similarityis also adopted to control the message delivery overheadsOn one hand since the node with high destination similarityhas more opportunities to encounter the destination nodethe message carrier node will choose these nodes as relayingnodes On the other hand the message carrier node will notreplicate themessage to the nodes with high carrier similaritybecause these nodes may have the mobility pattern similar tothe message carrier node

Secondly inspired by the law of diminishing marginalutility in economics we propose a buffer managementscheme for the routing algorithm [8] This law shows that asa user increases consumption of a product there is a declinein the marginal utility that user derives from consumingeach additional unit of that product Thus it is not goodto generate excessive redundant copies for a single messagebecause this will occupy very large buffer space of nodes andwill restrict the opportunity of other messages to access thebuffer space when buffer overflow happens In contrast inour proposed scheme we estimate the status of a messagefor example the total number of its copies in the network andits dissemination speed and perform buffer replacement andscheduling accordinglyMessages that have a larger estimatednumber of copies and faster dissemination speed are droppedprior to other messages in the event of buffer overflow orare forwarded after other messages In this way the proposedsolution provides better fairness among messages and henceimproves the overall performance

Themajor contributions of this paper are listed as follows

(i) A novel similarity-based routing (SBR) algorithm isproposed To the best of our knowledge no previousstudies have exploited the carrier similarity in design-ing routing protocols in DTNs

(ii) A buffer management scheme including messagereplacement and scheduling is proposed for therouting algorithm

(iii) Extensive simulation experiments are conducted onmap-based mobility model and random waypoint(RWP) mobility model The results demonstrate thatthe proposed routing algorithm combined with thebuffer management scheme can achieve the higherdelivery ratio and has a relatively lower overhead ratiocompared to other routing schemes

The rest of this paper is organized as follows Section 2reviews the previous work on routing and buffer manage-ment in DTNs Section 3 presents a similarity-based routingalgorithm Section 4 proposes a buffer management schemebased on the message transmission status In Section 5the routing algorithm with buffer management scheme isevaluated through simulation experiments Finally Section 6concludes the paper

2 Related Work

This section overviews state-of-the-art DTN routing proto-cols and buffer management issues

21 Single-Copy Routing Direct transmission is the basicapproach in single-copy routing The source node sendsthe message to the destination node directly Predict andrelay (PER) routing [9] used the probability distribution offuture contact times to choose the next relay node Recentlysocial characters have been explored to design routing proto-cols in DTNs MobySpace routing [10] constructed a high-dimensional Euclidean space upon mobility patterns Thenodes that have themobility pattern similar to the destinationare appropriate to act as relay nodes SimBet [11] routingexploited ego-centric centrality and its social similarity tochoose the better carriers for the final destination In [12]social selfishness was introduced into DTN routing anda social selfishness aware routing (SSAR) algorithm wasproposed for the collaborative environment Whatever therouting metric used however single-copy routing facesthe reliability problem that the message is not sent to itsdestination when the single message copy is discarded by therelay node

22 Multiple-Copy Routing The multiple-copy routing pro-tocols inject multiple copies of a message to provide reliabledelivery As one of the earliest DTN routing protocols theepidemic routing [13] floods the message throughout thenetwork and thus greatly increases themessage delivery ratioHowever the node resources such as buffer space band-width and energy seriously restrain the epidemic routingperformance To address this problemmany approaches havebeen adopted to reduce the overheads and improve the overallnetwork performance Based on the method of controllingthe overheads the multiple-copy routing schemes can befurther separated into three classes coding-based [14ndash17]quota-based [18ndash20] and utility-based [21ndash24]

In the coding-based algorithms a message is dividedinto a set of code blocks Destination node can reconstructthe original message when it receives a sufficiently largenumber of code blocks In [14] Wang et al combined erasurecoding and replication-based routing schemes to increase thenetwork throughput Tsapeli and Tsaoussidis [15] combinedthe probabilistic routing with erasure coding to enhance therobustness of the erasure-coding based forwarding in worst-case delays and small-delay scenarios In [17] Altman etal exploited linear block-codes and rateless random linearcoding to solve the problem of optimal transmission and

International Journal of Distributed Sensor Networks 3

scheduling policies with two-hop routing undermemory andenergy constraints

In the quota-based routing schemes the fixed numberof message copies is inserted into the network Thus theconstant overhead is maintained Spyropoulos et al [18] pro-posed two different methods that is Source Spray and Wait(SSampW) and Binary Spray and Wait (BSampW) for allocatingmessage copies In [19] Nelson et al used the encounter-basedmetric for optimization ofmessage allocation Elwhishiet al [20] proposed a self-adaptive contention aware routingprotocol SAURP which makes forwarding decision based onthe utility function value of the encountered node regardingthe destination and the number of message copy tokens

In the utility-based routing schemes each nodemaintainsa defined utility for destination The utility can be a functionof encounter history between nodes node resources and soforth Lindgren et al [21] used the past encounters to predictthe probability of future encounters Grasic et al [22] madea little modification to the routing metric calculations inProphet which can improve its performance Ramanathan etal [23] presented a variant of epidemic routing PREP whichprioritizes messages based on costs to destination sourceand expiry time to decide whether the message is deleted orreserved when it faced the situation of insufficient resourcesIn [24] Balasubramanian et al treated DTN routing as aproblem of resource allocation and proposed a heuristicapproach tomaximize the specified performancemetrics butit requires high computation cost

23 Buffer Management Generally since the main reasonof message loss is the buffer overflow [25ndash27] variouscongestion control schemes are proposed [28ndash30] In DTNhow to design buffer management schemes becomes veryimportant for the multiple-copy routing performance

Some buffer management strategies can be performedindependent of the underlying routing algorithm In theEpidemic-IMMUNE routing protocol [31] three buffermanagement strategies including drop-tail drop-head andsource-prioritized drop-head were examined It was shownthat appropriate buffer management schemes have greateffects on delivery performance In [32] Krifa et al proposeda distributed algorithm based on the estimated global infor-mation to optimize the delivery ratio and the average latency

Some other buffer management strategies have beenproposed for specific routing algorithms In [33] Lindgrenand Phanse used the delivery predictability metric definedin Prophet routing protocol to decide which message isforwarded first They evaluated various combinations ofqueuing policies and scheduling strategies for Prophet proto-col In [34] Erramilli and Crovella designed scheduling andreplacement algorithms based on the message priority thatis defined in delegation forwarding routing algorithm [35]In [36] Rashid et al proposed a buffer management policywhich takes message sizes into account

Compared to the above solutions our approach integratesboth routing protocol and buffer management strategy Theappropriate relay nodes are chosen by considering the carriersimilarity and destination similarity at the same time More-over the messages are replaced and scheduled based on the

number of copies and dissemination speed In this way ourapproach achieves higher delivery ratio and lower overheads

3 Similarity-Based Routing Algorithm

In this section the system model is presented The basicdefinitions and notations that will be used throughout therest of the paper are also introducedThenmobility similaritymetrics used for routing decision are defined The design ofsimilarity-based routing algorithm is presented in the end

31 System Model The network consists of 119873 mobile nodeseach ofwhich has a unique ID and belongs to one communityEach node in the network has the same buffer size andthe same transmission range The probability of moving toother communities from the local community is the sameThe global knowledge about nodesrsquo mobility is unknown toevery node in the beginning Every message generated in thenetwork has a Time-to-Live (TTL) value The source nodeand relay nodeswill dropmessage and their copies when theirTTL expires For a given message there are one source nodeone destination node and (119873 minus 2) intermediate nodes Thesource node can delivermessage to the destination directly orthrough intermediate nodes Two nodes exchange messageswhen they are within communication range of each otherWe assume that the bandwidth is large enough to transmitall messages

The contact time is defined as the time interval in whichtwo nodes are within their communication range The inter-contact time is defined as the time interval in which a nodepair is not within its communication rangeThe interdroppingtime is defined as the time interval in which a single messageor its copies are dropped

Because the transmission time is neglectable compared tothe intercontact time the transmission time is often ignoredin other studies Some studies indicated that many popularmobility models like random waypoint random walk andcommunity-based model have such a property where theintercontact time of mobility model is exponentially dis-tributed or has exponential tail [37 38]We also consider thatthe intercontact time follows exponential distributions Theinterdropping time of messages is mutually independent andidentically distributed random variables with an exponentialdistribution

32 Mobility Similarity Metrics Nodes with similar charac-teristics of the mobility pattern are discovered in designingrouting metrics in DTNs [10 11] Some previous studieshave exploited mobility similarity as routing metric butthey only considered the destination similarity It is arguedthat the nodes having the similar mobility pattern withthe destination can enhance the delivery ratio and reducethe communication cost However from the viewpoint ofmessage carrier node if the encountered nodes have similarmobility pattern with the message carrier node then theywould not be good relaying nodes Thus we propose thecarrier similarity to control the overhead of multiple messagecopies

4 International Journal of Distributed Sensor Networks

Carrier similarity is measured as the proportion of thesame history encountered nodes between themessage carriernode and the contact node If a node has high similarity withthe message carrier node it will have similar mobility scopeThe node will not be appropriate for diffusing message faraway The message carrier node will hold the messages byitself In contrast if a node has low similaritywith themessagecarrier node it will be capable of disseminating messages faraway Thus the node is a good potential relaying node Let 119878

119894

and 119878119895denote the set of encounters experienced by node 119894 and

node 119895 within 119879 respectively When node 119894 contacts node 119895the carrier similarity Sim119904

119894 calculated by node 119894 is defined as

follows

Sim119904119894=

10038161003816100381610038161003816119878119894cap 119878119895

100381610038161003816100381610038161003816100381610038161003816119878119894

1003816100381610038161003816

(1)

Destination similarity measures the number of the samenodes encountered by a node and the destination node If anode has high similarity with the destination it may have ahigher chance to meet the destination For a given node 119894 thedestination similarity is defined below

Sim119889119894=

1003816100381610038161003816119878119894 cap 119878119889

1003816100381610038161003816 (2)

where 119878119894and 119878

119889are the set of the encounter nodes expe-

rienced by node 119894 and the destination node within 119879respectively

33 Similarity-Based Routing Algorithm Here we proposethe similarity-based routing algorithm with the aim ofbringing a concise concept of behavior similarity into DTNrouting Similarity-based routing combines carrier and des-tination similarity to make routing decisions There are twointuitions behind this algorithm Carrier similarity couldcontrol the number of flooding message copies and avoidmissing the good relay node Destination similarity couldimprove forwarding efficiency

Each node maintains an encounter node vector (nv) Anentry in the nv is composed of two items including node IDand encounter timeThenode can obtain contact informationin regard to nodes that it has encountered and that it has notyet encountered When a node encounters other nodes thenode will exchange the nv If the encounter node already hasan entry in the nv the node will only modify the encountertime Otherwise a new entry will be inserted into the nvThenode will refresh its nv in a given time 119879 If the encountertime of a node exceeds the time 119879 the entry of the node willbe removed from the nv

Node contacts are represented as an 119899 times 119899 symmetricmatrix where 119899 is the number of contacts encountered by agiven node If there is a contact between node 119894 and node 119895the corresponding element in the matrix is set to 1 otherwiseto 0 The count of nonzero equivalent row entries in thematrix represents the number of common neighbors betweennode 119894 and node 119895 The intermediate nodes may obtain theuseful information in regard to the destination node viadirect encountered nodes The destination node may be anindirect node Then the similarity with indirect encounter

Upon meeting node jif hasMsgsForDest(j) == true thendeliverMsgs(j)

end ifexchangeEncountersVector(j nv)updateSimilarity( )exchangeSummaryVector( )if Sim119904

119894lt Similarity119904TH then

for all messages unknown to node j doif Sim119889

119895ge Sim119889

119894then

replicateMsgs(j messages)end if

end forelsefor all messages unknown to node j doif Sim119889

119895gt Sim119889

119894then

forwardMsgs(j messages)end if

end forend ifUpon reception messagem from node jupdateSummaryVector( )

Algorithm 1 Similarity-based routing algorithm SBR

nodes is also needed to be evaluated Node contacts withdirect encountered nodes and indirect encountered nodes arerepresented as an 119899 times 119896 matrix where 119899 is the number ofdirect encountered nodes and 119896 is the number of indirectencountered nodes More details of the calculation can befound in [11]

Each node also maintains a summary vector An entryin summary vector is composed of four items includingmessage ID replication number message TTL value andinitial message TTL value The summary vector is used toupdate the forwarding status of a message It is exchangedwhen two nodes are encountered When a node receives anew message a new entry is added in the summary vectorWhen the node discards a message the corresponding entryin the summary vector remains in the node until the TTLvalue of that message expires

Because the sizes of the encounter node vector and thesummary vector are very small compared to a message theoverheads of the encounter vector and the summary vectorare not considered

The routing algorithm is outlined inAlgorithm 1 It repre-sents the communication process whenmessage carrier node119894 meets the potential relaying node 119895 They exchange someinformation of their encounter history used for computingroutingmetrics like Epidemic Node 119894 calculates the similaritywith node 119895

If the carrier similarity utility between node 119894 andnode 119895 is below the predefined carrier similarity thresholdSimilarity119904TH it means that node 119895 has the ability to spreadthe message to more unfamiliar nodes If the threshold istoo small messages are hard to be sent to other nodes Ifthe threshold is too large excessive distribution of messagescannot be controlled Therefore Similarity119904TH is set as 05 to

International Journal of Distributed Sensor Networks 5

achieve a good trade-off between the overhead and deliveryratio For all the unknown messages carried by node 119894 whenthe destination of messages has higher similarity with node 119895

than it has with node 119894 node 119894 will replicate these messagesto node 119895 If the carrier similarity utility between node 119894

and node 119895 is greater than the predefined carrier similaritythreshold it means that node 119894 and node 119895 have similarmobility scope There is no need to increase the numberof message copies in the networks Then node 119894 will onlyforward these messages to node 119895 when the latter has a highdestination similarity utility

4 Buffer Management Scheme

Although multiple-copy routing can improve the probabilityof delivery rate it also inevitably brings more traffic intothe network When the node buffer overflows the networkperformance decreases sharply To overcome this problemwe propose a buffer management approach which exploitsmessages transmission status in the networks to decide thepriority of message replacement and scheduling

41 Message Transmission Analysis The dissemination ofmessages in the network is modeled from the perspective ofan individual message 119898 The nodes that hold message 119898 orits copy in the networks are called infected nodes other nodesare called uninfected nodes Let 119909 and 119904 denote the number ofinfected nodes and uninfected nodes respectively at a giventime Thus the total number of the network nodes can bewritten as

119873 = 119904+119909 (3)

Let 120582 represent the encounter rate between the nodes Eachinfected nodewillmeet120582119904119873 uninfected nodes in a unit timeThen the total number of infected nodes increases by 120582119904119909119873

in a unit time Let120583 denote the rate of droppingmessagesThenumber of nodes whose status is changed from infected touninfected is 120583119909 Then the increasing rate of infected nodesis

119889119909

119889119905=

120582119904119909

119873minus120583119909 (4)

Messages will be discarded when two encountered nodesexchange message copies and buffer overflow occurs Underthe condition that the encounter rate of nodes is greater thanthe dropping rate of messages that is 120582 gt 120583 let

120588 =120582

120583 (5)

Combining (3) (4) and (5) together yields

119889119909

119889119905= minus 120582119909 [

119909

119873minus(1minus

1120588)] (6)

The number of infected nodes can be expressed as

119909 = 119873(1minus1120588)

11 + [119873 (1 minus 1120588) minus 1] 119890minus120582(1minus1120588)119905

(7)

We can obtain from (5) and (7) that the increasing rate of 119909depends on its initial value Because the initial value of 119909 is1 119909 grows faster in the beginning stage of disseminationThelimit value increases as 120588 increases and is given by

119909 (infin) = 119873(1minus1120588) (8)

Let a utility function 119880(119909) model the delivery probabilityfor each single message 119880(119909) is an increasing function ofparameter 119909 In other words the greater the number of nodesreceiving the samemessage the higher delivery probability ofthe message This implies the following requirement on thederivative of the function 119880(119909)

119889119880 (119909)

119889119909ge 0 (9)

According to the law of diminishing marginal utility we canobtain the equation below

lim119909rarr119873

119889119880 (119909)

119889119909= 0 (10)

Equation (10) reflects the phenomenon that the improvementof the delivery probability is vanishing when a high deliveryprobability is reached that is a feasible assumption for generalcases

Note that in the case of buffer management the numberof infected nodes is not greater than the total number of nodesin the network Therefore there is an upper bound to theprobability for each single message For this reason we get

lim119909rarr119873

119880 (119909) = 1 (11)

It is often supposed that a utility function has some propertiesof regularity for example continuous differentiability at leastpiecewise When these properties are applied to (9) (10) and(11) we can have

exist119888

10158401015840

119880 (119909) lt 0 forall119909 ge 119888 (12)

Formula (12) implies that the concavity of119880(119909) at least for119909 is greater than a given valueTherefore when amessage hasfewer copies in the network119880(119909)has the properties of convexfunction From the perspective of the whole network if sucha message is dropped in the network the decreased utility ofthis message is larger than the increased utility when a nodeaccommodates a message with a large number of copies Onthe contrary when a message has a large number of copies inthe network 119880(119909) has the properties of concave function Ifsuch a message is dropped the decreased utility is less thanthe increased utility when a node accepts a message with fewcopies

From the above analysis we can obtain thatmessageswitha large number of copies in the network have much moretransfer opportunities These messages will have relativelyhigher probability to reach their destination nodes Whenwe control the increasing number of these messages thelost utility is limited at a low level For a single message it

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

2 International Journal of Distributed Sensor Networks

part with external traffic when acting as a relaying nodeThus how to manage buffer space significantly affects theperformance of multiple-copy routing protocols especiallyin the environments where intermittent connectivity andlong latency require the data to be stored for a long periodthroughout the network

In this paper we firstly propose a similarity-based routingalgorithm in DTNs Two mobility similarity metrics areadopted that is destination similarity and carrier similarityto choose the next relay nodeThe destination similarity refersto the mobility similarity between an intermediate node andthe destination The carrier similarity refers to the mobilitysimilarity between the message carrier node and its encoun-tered node In addition to using the destination similarity toimprove message delivery probability the carrier similarityis also adopted to control the message delivery overheadsOn one hand since the node with high destination similarityhas more opportunities to encounter the destination nodethe message carrier node will choose these nodes as relayingnodes On the other hand the message carrier node will notreplicate themessage to the nodes with high carrier similaritybecause these nodes may have the mobility pattern similar tothe message carrier node

Secondly inspired by the law of diminishing marginalutility in economics we propose a buffer managementscheme for the routing algorithm [8] This law shows that asa user increases consumption of a product there is a declinein the marginal utility that user derives from consumingeach additional unit of that product Thus it is not goodto generate excessive redundant copies for a single messagebecause this will occupy very large buffer space of nodes andwill restrict the opportunity of other messages to access thebuffer space when buffer overflow happens In contrast inour proposed scheme we estimate the status of a messagefor example the total number of its copies in the network andits dissemination speed and perform buffer replacement andscheduling accordinglyMessages that have a larger estimatednumber of copies and faster dissemination speed are droppedprior to other messages in the event of buffer overflow orare forwarded after other messages In this way the proposedsolution provides better fairness among messages and henceimproves the overall performance

Themajor contributions of this paper are listed as follows

(i) A novel similarity-based routing (SBR) algorithm isproposed To the best of our knowledge no previousstudies have exploited the carrier similarity in design-ing routing protocols in DTNs

(ii) A buffer management scheme including messagereplacement and scheduling is proposed for therouting algorithm

(iii) Extensive simulation experiments are conducted onmap-based mobility model and random waypoint(RWP) mobility model The results demonstrate thatthe proposed routing algorithm combined with thebuffer management scheme can achieve the higherdelivery ratio and has a relatively lower overhead ratiocompared to other routing schemes

The rest of this paper is organized as follows Section 2reviews the previous work on routing and buffer manage-ment in DTNs Section 3 presents a similarity-based routingalgorithm Section 4 proposes a buffer management schemebased on the message transmission status In Section 5the routing algorithm with buffer management scheme isevaluated through simulation experiments Finally Section 6concludes the paper

2 Related Work

This section overviews state-of-the-art DTN routing proto-cols and buffer management issues

21 Single-Copy Routing Direct transmission is the basicapproach in single-copy routing The source node sendsthe message to the destination node directly Predict andrelay (PER) routing [9] used the probability distribution offuture contact times to choose the next relay node Recentlysocial characters have been explored to design routing proto-cols in DTNs MobySpace routing [10] constructed a high-dimensional Euclidean space upon mobility patterns Thenodes that have themobility pattern similar to the destinationare appropriate to act as relay nodes SimBet [11] routingexploited ego-centric centrality and its social similarity tochoose the better carriers for the final destination In [12]social selfishness was introduced into DTN routing anda social selfishness aware routing (SSAR) algorithm wasproposed for the collaborative environment Whatever therouting metric used however single-copy routing facesthe reliability problem that the message is not sent to itsdestination when the single message copy is discarded by therelay node

22 Multiple-Copy Routing The multiple-copy routing pro-tocols inject multiple copies of a message to provide reliabledelivery As one of the earliest DTN routing protocols theepidemic routing [13] floods the message throughout thenetwork and thus greatly increases themessage delivery ratioHowever the node resources such as buffer space band-width and energy seriously restrain the epidemic routingperformance To address this problemmany approaches havebeen adopted to reduce the overheads and improve the overallnetwork performance Based on the method of controllingthe overheads the multiple-copy routing schemes can befurther separated into three classes coding-based [14ndash17]quota-based [18ndash20] and utility-based [21ndash24]

In the coding-based algorithms a message is dividedinto a set of code blocks Destination node can reconstructthe original message when it receives a sufficiently largenumber of code blocks In [14] Wang et al combined erasurecoding and replication-based routing schemes to increase thenetwork throughput Tsapeli and Tsaoussidis [15] combinedthe probabilistic routing with erasure coding to enhance therobustness of the erasure-coding based forwarding in worst-case delays and small-delay scenarios In [17] Altman etal exploited linear block-codes and rateless random linearcoding to solve the problem of optimal transmission and

International Journal of Distributed Sensor Networks 3

scheduling policies with two-hop routing undermemory andenergy constraints

In the quota-based routing schemes the fixed numberof message copies is inserted into the network Thus theconstant overhead is maintained Spyropoulos et al [18] pro-posed two different methods that is Source Spray and Wait(SSampW) and Binary Spray and Wait (BSampW) for allocatingmessage copies In [19] Nelson et al used the encounter-basedmetric for optimization ofmessage allocation Elwhishiet al [20] proposed a self-adaptive contention aware routingprotocol SAURP which makes forwarding decision based onthe utility function value of the encountered node regardingthe destination and the number of message copy tokens

In the utility-based routing schemes each nodemaintainsa defined utility for destination The utility can be a functionof encounter history between nodes node resources and soforth Lindgren et al [21] used the past encounters to predictthe probability of future encounters Grasic et al [22] madea little modification to the routing metric calculations inProphet which can improve its performance Ramanathan etal [23] presented a variant of epidemic routing PREP whichprioritizes messages based on costs to destination sourceand expiry time to decide whether the message is deleted orreserved when it faced the situation of insufficient resourcesIn [24] Balasubramanian et al treated DTN routing as aproblem of resource allocation and proposed a heuristicapproach tomaximize the specified performancemetrics butit requires high computation cost

23 Buffer Management Generally since the main reasonof message loss is the buffer overflow [25ndash27] variouscongestion control schemes are proposed [28ndash30] In DTNhow to design buffer management schemes becomes veryimportant for the multiple-copy routing performance

Some buffer management strategies can be performedindependent of the underlying routing algorithm In theEpidemic-IMMUNE routing protocol [31] three buffermanagement strategies including drop-tail drop-head andsource-prioritized drop-head were examined It was shownthat appropriate buffer management schemes have greateffects on delivery performance In [32] Krifa et al proposeda distributed algorithm based on the estimated global infor-mation to optimize the delivery ratio and the average latency

Some other buffer management strategies have beenproposed for specific routing algorithms In [33] Lindgrenand Phanse used the delivery predictability metric definedin Prophet routing protocol to decide which message isforwarded first They evaluated various combinations ofqueuing policies and scheduling strategies for Prophet proto-col In [34] Erramilli and Crovella designed scheduling andreplacement algorithms based on the message priority thatis defined in delegation forwarding routing algorithm [35]In [36] Rashid et al proposed a buffer management policywhich takes message sizes into account

Compared to the above solutions our approach integratesboth routing protocol and buffer management strategy Theappropriate relay nodes are chosen by considering the carriersimilarity and destination similarity at the same time More-over the messages are replaced and scheduled based on the

number of copies and dissemination speed In this way ourapproach achieves higher delivery ratio and lower overheads

3 Similarity-Based Routing Algorithm

In this section the system model is presented The basicdefinitions and notations that will be used throughout therest of the paper are also introducedThenmobility similaritymetrics used for routing decision are defined The design ofsimilarity-based routing algorithm is presented in the end

31 System Model The network consists of 119873 mobile nodeseach ofwhich has a unique ID and belongs to one communityEach node in the network has the same buffer size andthe same transmission range The probability of moving toother communities from the local community is the sameThe global knowledge about nodesrsquo mobility is unknown toevery node in the beginning Every message generated in thenetwork has a Time-to-Live (TTL) value The source nodeand relay nodeswill dropmessage and their copies when theirTTL expires For a given message there are one source nodeone destination node and (119873 minus 2) intermediate nodes Thesource node can delivermessage to the destination directly orthrough intermediate nodes Two nodes exchange messageswhen they are within communication range of each otherWe assume that the bandwidth is large enough to transmitall messages

The contact time is defined as the time interval in whichtwo nodes are within their communication range The inter-contact time is defined as the time interval in which a nodepair is not within its communication rangeThe interdroppingtime is defined as the time interval in which a single messageor its copies are dropped

Because the transmission time is neglectable compared tothe intercontact time the transmission time is often ignoredin other studies Some studies indicated that many popularmobility models like random waypoint random walk andcommunity-based model have such a property where theintercontact time of mobility model is exponentially dis-tributed or has exponential tail [37 38]We also consider thatthe intercontact time follows exponential distributions Theinterdropping time of messages is mutually independent andidentically distributed random variables with an exponentialdistribution

32 Mobility Similarity Metrics Nodes with similar charac-teristics of the mobility pattern are discovered in designingrouting metrics in DTNs [10 11] Some previous studieshave exploited mobility similarity as routing metric butthey only considered the destination similarity It is arguedthat the nodes having the similar mobility pattern withthe destination can enhance the delivery ratio and reducethe communication cost However from the viewpoint ofmessage carrier node if the encountered nodes have similarmobility pattern with the message carrier node then theywould not be good relaying nodes Thus we propose thecarrier similarity to control the overhead of multiple messagecopies

4 International Journal of Distributed Sensor Networks

Carrier similarity is measured as the proportion of thesame history encountered nodes between themessage carriernode and the contact node If a node has high similarity withthe message carrier node it will have similar mobility scopeThe node will not be appropriate for diffusing message faraway The message carrier node will hold the messages byitself In contrast if a node has low similaritywith themessagecarrier node it will be capable of disseminating messages faraway Thus the node is a good potential relaying node Let 119878

119894

and 119878119895denote the set of encounters experienced by node 119894 and

node 119895 within 119879 respectively When node 119894 contacts node 119895the carrier similarity Sim119904

119894 calculated by node 119894 is defined as

follows

Sim119904119894=

10038161003816100381610038161003816119878119894cap 119878119895

100381610038161003816100381610038161003816100381610038161003816119878119894

1003816100381610038161003816

(1)

Destination similarity measures the number of the samenodes encountered by a node and the destination node If anode has high similarity with the destination it may have ahigher chance to meet the destination For a given node 119894 thedestination similarity is defined below

Sim119889119894=

1003816100381610038161003816119878119894 cap 119878119889

1003816100381610038161003816 (2)

where 119878119894and 119878

119889are the set of the encounter nodes expe-

rienced by node 119894 and the destination node within 119879respectively

33 Similarity-Based Routing Algorithm Here we proposethe similarity-based routing algorithm with the aim ofbringing a concise concept of behavior similarity into DTNrouting Similarity-based routing combines carrier and des-tination similarity to make routing decisions There are twointuitions behind this algorithm Carrier similarity couldcontrol the number of flooding message copies and avoidmissing the good relay node Destination similarity couldimprove forwarding efficiency

Each node maintains an encounter node vector (nv) Anentry in the nv is composed of two items including node IDand encounter timeThenode can obtain contact informationin regard to nodes that it has encountered and that it has notyet encountered When a node encounters other nodes thenode will exchange the nv If the encounter node already hasan entry in the nv the node will only modify the encountertime Otherwise a new entry will be inserted into the nvThenode will refresh its nv in a given time 119879 If the encountertime of a node exceeds the time 119879 the entry of the node willbe removed from the nv

Node contacts are represented as an 119899 times 119899 symmetricmatrix where 119899 is the number of contacts encountered by agiven node If there is a contact between node 119894 and node 119895the corresponding element in the matrix is set to 1 otherwiseto 0 The count of nonzero equivalent row entries in thematrix represents the number of common neighbors betweennode 119894 and node 119895 The intermediate nodes may obtain theuseful information in regard to the destination node viadirect encountered nodes The destination node may be anindirect node Then the similarity with indirect encounter

Upon meeting node jif hasMsgsForDest(j) == true thendeliverMsgs(j)

end ifexchangeEncountersVector(j nv)updateSimilarity( )exchangeSummaryVector( )if Sim119904

119894lt Similarity119904TH then

for all messages unknown to node j doif Sim119889

119895ge Sim119889

119894then

replicateMsgs(j messages)end if

end forelsefor all messages unknown to node j doif Sim119889

119895gt Sim119889

119894then

forwardMsgs(j messages)end if

end forend ifUpon reception messagem from node jupdateSummaryVector( )

Algorithm 1 Similarity-based routing algorithm SBR

nodes is also needed to be evaluated Node contacts withdirect encountered nodes and indirect encountered nodes arerepresented as an 119899 times 119896 matrix where 119899 is the number ofdirect encountered nodes and 119896 is the number of indirectencountered nodes More details of the calculation can befound in [11]

Each node also maintains a summary vector An entryin summary vector is composed of four items includingmessage ID replication number message TTL value andinitial message TTL value The summary vector is used toupdate the forwarding status of a message It is exchangedwhen two nodes are encountered When a node receives anew message a new entry is added in the summary vectorWhen the node discards a message the corresponding entryin the summary vector remains in the node until the TTLvalue of that message expires

Because the sizes of the encounter node vector and thesummary vector are very small compared to a message theoverheads of the encounter vector and the summary vectorare not considered

The routing algorithm is outlined inAlgorithm 1 It repre-sents the communication process whenmessage carrier node119894 meets the potential relaying node 119895 They exchange someinformation of their encounter history used for computingroutingmetrics like Epidemic Node 119894 calculates the similaritywith node 119895

If the carrier similarity utility between node 119894 andnode 119895 is below the predefined carrier similarity thresholdSimilarity119904TH it means that node 119895 has the ability to spreadthe message to more unfamiliar nodes If the threshold istoo small messages are hard to be sent to other nodes Ifthe threshold is too large excessive distribution of messagescannot be controlled Therefore Similarity119904TH is set as 05 to

International Journal of Distributed Sensor Networks 5

achieve a good trade-off between the overhead and deliveryratio For all the unknown messages carried by node 119894 whenthe destination of messages has higher similarity with node 119895

than it has with node 119894 node 119894 will replicate these messagesto node 119895 If the carrier similarity utility between node 119894

and node 119895 is greater than the predefined carrier similaritythreshold it means that node 119894 and node 119895 have similarmobility scope There is no need to increase the numberof message copies in the networks Then node 119894 will onlyforward these messages to node 119895 when the latter has a highdestination similarity utility

4 Buffer Management Scheme

Although multiple-copy routing can improve the probabilityof delivery rate it also inevitably brings more traffic intothe network When the node buffer overflows the networkperformance decreases sharply To overcome this problemwe propose a buffer management approach which exploitsmessages transmission status in the networks to decide thepriority of message replacement and scheduling

41 Message Transmission Analysis The dissemination ofmessages in the network is modeled from the perspective ofan individual message 119898 The nodes that hold message 119898 orits copy in the networks are called infected nodes other nodesare called uninfected nodes Let 119909 and 119904 denote the number ofinfected nodes and uninfected nodes respectively at a giventime Thus the total number of the network nodes can bewritten as

119873 = 119904+119909 (3)

Let 120582 represent the encounter rate between the nodes Eachinfected nodewillmeet120582119904119873 uninfected nodes in a unit timeThen the total number of infected nodes increases by 120582119904119909119873

in a unit time Let120583 denote the rate of droppingmessagesThenumber of nodes whose status is changed from infected touninfected is 120583119909 Then the increasing rate of infected nodesis

119889119909

119889119905=

120582119904119909

119873minus120583119909 (4)

Messages will be discarded when two encountered nodesexchange message copies and buffer overflow occurs Underthe condition that the encounter rate of nodes is greater thanthe dropping rate of messages that is 120582 gt 120583 let

120588 =120582

120583 (5)

Combining (3) (4) and (5) together yields

119889119909

119889119905= minus 120582119909 [

119909

119873minus(1minus

1120588)] (6)

The number of infected nodes can be expressed as

119909 = 119873(1minus1120588)

11 + [119873 (1 minus 1120588) minus 1] 119890minus120582(1minus1120588)119905

(7)

We can obtain from (5) and (7) that the increasing rate of 119909depends on its initial value Because the initial value of 119909 is1 119909 grows faster in the beginning stage of disseminationThelimit value increases as 120588 increases and is given by

119909 (infin) = 119873(1minus1120588) (8)

Let a utility function 119880(119909) model the delivery probabilityfor each single message 119880(119909) is an increasing function ofparameter 119909 In other words the greater the number of nodesreceiving the samemessage the higher delivery probability ofthe message This implies the following requirement on thederivative of the function 119880(119909)

119889119880 (119909)

119889119909ge 0 (9)

According to the law of diminishing marginal utility we canobtain the equation below

lim119909rarr119873

119889119880 (119909)

119889119909= 0 (10)

Equation (10) reflects the phenomenon that the improvementof the delivery probability is vanishing when a high deliveryprobability is reached that is a feasible assumption for generalcases

Note that in the case of buffer management the numberof infected nodes is not greater than the total number of nodesin the network Therefore there is an upper bound to theprobability for each single message For this reason we get

lim119909rarr119873

119880 (119909) = 1 (11)

It is often supposed that a utility function has some propertiesof regularity for example continuous differentiability at leastpiecewise When these properties are applied to (9) (10) and(11) we can have

exist119888

10158401015840

119880 (119909) lt 0 forall119909 ge 119888 (12)

Formula (12) implies that the concavity of119880(119909) at least for119909 is greater than a given valueTherefore when amessage hasfewer copies in the network119880(119909)has the properties of convexfunction From the perspective of the whole network if sucha message is dropped in the network the decreased utility ofthis message is larger than the increased utility when a nodeaccommodates a message with a large number of copies Onthe contrary when a message has a large number of copies inthe network 119880(119909) has the properties of concave function Ifsuch a message is dropped the decreased utility is less thanthe increased utility when a node accepts a message with fewcopies

From the above analysis we can obtain thatmessageswitha large number of copies in the network have much moretransfer opportunities These messages will have relativelyhigher probability to reach their destination nodes Whenwe control the increasing number of these messages thelost utility is limited at a low level For a single message it

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 3

scheduling policies with two-hop routing undermemory andenergy constraints

In the quota-based routing schemes the fixed numberof message copies is inserted into the network Thus theconstant overhead is maintained Spyropoulos et al [18] pro-posed two different methods that is Source Spray and Wait(SSampW) and Binary Spray and Wait (BSampW) for allocatingmessage copies In [19] Nelson et al used the encounter-basedmetric for optimization ofmessage allocation Elwhishiet al [20] proposed a self-adaptive contention aware routingprotocol SAURP which makes forwarding decision based onthe utility function value of the encountered node regardingthe destination and the number of message copy tokens

In the utility-based routing schemes each nodemaintainsa defined utility for destination The utility can be a functionof encounter history between nodes node resources and soforth Lindgren et al [21] used the past encounters to predictthe probability of future encounters Grasic et al [22] madea little modification to the routing metric calculations inProphet which can improve its performance Ramanathan etal [23] presented a variant of epidemic routing PREP whichprioritizes messages based on costs to destination sourceand expiry time to decide whether the message is deleted orreserved when it faced the situation of insufficient resourcesIn [24] Balasubramanian et al treated DTN routing as aproblem of resource allocation and proposed a heuristicapproach tomaximize the specified performancemetrics butit requires high computation cost

23 Buffer Management Generally since the main reasonof message loss is the buffer overflow [25ndash27] variouscongestion control schemes are proposed [28ndash30] In DTNhow to design buffer management schemes becomes veryimportant for the multiple-copy routing performance

Some buffer management strategies can be performedindependent of the underlying routing algorithm In theEpidemic-IMMUNE routing protocol [31] three buffermanagement strategies including drop-tail drop-head andsource-prioritized drop-head were examined It was shownthat appropriate buffer management schemes have greateffects on delivery performance In [32] Krifa et al proposeda distributed algorithm based on the estimated global infor-mation to optimize the delivery ratio and the average latency

Some other buffer management strategies have beenproposed for specific routing algorithms In [33] Lindgrenand Phanse used the delivery predictability metric definedin Prophet routing protocol to decide which message isforwarded first They evaluated various combinations ofqueuing policies and scheduling strategies for Prophet proto-col In [34] Erramilli and Crovella designed scheduling andreplacement algorithms based on the message priority thatis defined in delegation forwarding routing algorithm [35]In [36] Rashid et al proposed a buffer management policywhich takes message sizes into account

Compared to the above solutions our approach integratesboth routing protocol and buffer management strategy Theappropriate relay nodes are chosen by considering the carriersimilarity and destination similarity at the same time More-over the messages are replaced and scheduled based on the

number of copies and dissemination speed In this way ourapproach achieves higher delivery ratio and lower overheads

3 Similarity-Based Routing Algorithm

In this section the system model is presented The basicdefinitions and notations that will be used throughout therest of the paper are also introducedThenmobility similaritymetrics used for routing decision are defined The design ofsimilarity-based routing algorithm is presented in the end

31 System Model The network consists of 119873 mobile nodeseach ofwhich has a unique ID and belongs to one communityEach node in the network has the same buffer size andthe same transmission range The probability of moving toother communities from the local community is the sameThe global knowledge about nodesrsquo mobility is unknown toevery node in the beginning Every message generated in thenetwork has a Time-to-Live (TTL) value The source nodeand relay nodeswill dropmessage and their copies when theirTTL expires For a given message there are one source nodeone destination node and (119873 minus 2) intermediate nodes Thesource node can delivermessage to the destination directly orthrough intermediate nodes Two nodes exchange messageswhen they are within communication range of each otherWe assume that the bandwidth is large enough to transmitall messages

The contact time is defined as the time interval in whichtwo nodes are within their communication range The inter-contact time is defined as the time interval in which a nodepair is not within its communication rangeThe interdroppingtime is defined as the time interval in which a single messageor its copies are dropped

Because the transmission time is neglectable compared tothe intercontact time the transmission time is often ignoredin other studies Some studies indicated that many popularmobility models like random waypoint random walk andcommunity-based model have such a property where theintercontact time of mobility model is exponentially dis-tributed or has exponential tail [37 38]We also consider thatthe intercontact time follows exponential distributions Theinterdropping time of messages is mutually independent andidentically distributed random variables with an exponentialdistribution

32 Mobility Similarity Metrics Nodes with similar charac-teristics of the mobility pattern are discovered in designingrouting metrics in DTNs [10 11] Some previous studieshave exploited mobility similarity as routing metric butthey only considered the destination similarity It is arguedthat the nodes having the similar mobility pattern withthe destination can enhance the delivery ratio and reducethe communication cost However from the viewpoint ofmessage carrier node if the encountered nodes have similarmobility pattern with the message carrier node then theywould not be good relaying nodes Thus we propose thecarrier similarity to control the overhead of multiple messagecopies

4 International Journal of Distributed Sensor Networks

Carrier similarity is measured as the proportion of thesame history encountered nodes between themessage carriernode and the contact node If a node has high similarity withthe message carrier node it will have similar mobility scopeThe node will not be appropriate for diffusing message faraway The message carrier node will hold the messages byitself In contrast if a node has low similaritywith themessagecarrier node it will be capable of disseminating messages faraway Thus the node is a good potential relaying node Let 119878

119894

and 119878119895denote the set of encounters experienced by node 119894 and

node 119895 within 119879 respectively When node 119894 contacts node 119895the carrier similarity Sim119904

119894 calculated by node 119894 is defined as

follows

Sim119904119894=

10038161003816100381610038161003816119878119894cap 119878119895

100381610038161003816100381610038161003816100381610038161003816119878119894

1003816100381610038161003816

(1)

Destination similarity measures the number of the samenodes encountered by a node and the destination node If anode has high similarity with the destination it may have ahigher chance to meet the destination For a given node 119894 thedestination similarity is defined below

Sim119889119894=

1003816100381610038161003816119878119894 cap 119878119889

1003816100381610038161003816 (2)

where 119878119894and 119878

119889are the set of the encounter nodes expe-

rienced by node 119894 and the destination node within 119879respectively

33 Similarity-Based Routing Algorithm Here we proposethe similarity-based routing algorithm with the aim ofbringing a concise concept of behavior similarity into DTNrouting Similarity-based routing combines carrier and des-tination similarity to make routing decisions There are twointuitions behind this algorithm Carrier similarity couldcontrol the number of flooding message copies and avoidmissing the good relay node Destination similarity couldimprove forwarding efficiency

Each node maintains an encounter node vector (nv) Anentry in the nv is composed of two items including node IDand encounter timeThenode can obtain contact informationin regard to nodes that it has encountered and that it has notyet encountered When a node encounters other nodes thenode will exchange the nv If the encounter node already hasan entry in the nv the node will only modify the encountertime Otherwise a new entry will be inserted into the nvThenode will refresh its nv in a given time 119879 If the encountertime of a node exceeds the time 119879 the entry of the node willbe removed from the nv

Node contacts are represented as an 119899 times 119899 symmetricmatrix where 119899 is the number of contacts encountered by agiven node If there is a contact between node 119894 and node 119895the corresponding element in the matrix is set to 1 otherwiseto 0 The count of nonzero equivalent row entries in thematrix represents the number of common neighbors betweennode 119894 and node 119895 The intermediate nodes may obtain theuseful information in regard to the destination node viadirect encountered nodes The destination node may be anindirect node Then the similarity with indirect encounter

Upon meeting node jif hasMsgsForDest(j) == true thendeliverMsgs(j)

end ifexchangeEncountersVector(j nv)updateSimilarity( )exchangeSummaryVector( )if Sim119904

119894lt Similarity119904TH then

for all messages unknown to node j doif Sim119889

119895ge Sim119889

119894then

replicateMsgs(j messages)end if

end forelsefor all messages unknown to node j doif Sim119889

119895gt Sim119889

119894then

forwardMsgs(j messages)end if

end forend ifUpon reception messagem from node jupdateSummaryVector( )

Algorithm 1 Similarity-based routing algorithm SBR

nodes is also needed to be evaluated Node contacts withdirect encountered nodes and indirect encountered nodes arerepresented as an 119899 times 119896 matrix where 119899 is the number ofdirect encountered nodes and 119896 is the number of indirectencountered nodes More details of the calculation can befound in [11]

Each node also maintains a summary vector An entryin summary vector is composed of four items includingmessage ID replication number message TTL value andinitial message TTL value The summary vector is used toupdate the forwarding status of a message It is exchangedwhen two nodes are encountered When a node receives anew message a new entry is added in the summary vectorWhen the node discards a message the corresponding entryin the summary vector remains in the node until the TTLvalue of that message expires

Because the sizes of the encounter node vector and thesummary vector are very small compared to a message theoverheads of the encounter vector and the summary vectorare not considered

The routing algorithm is outlined inAlgorithm 1 It repre-sents the communication process whenmessage carrier node119894 meets the potential relaying node 119895 They exchange someinformation of their encounter history used for computingroutingmetrics like Epidemic Node 119894 calculates the similaritywith node 119895

If the carrier similarity utility between node 119894 andnode 119895 is below the predefined carrier similarity thresholdSimilarity119904TH it means that node 119895 has the ability to spreadthe message to more unfamiliar nodes If the threshold istoo small messages are hard to be sent to other nodes Ifthe threshold is too large excessive distribution of messagescannot be controlled Therefore Similarity119904TH is set as 05 to

International Journal of Distributed Sensor Networks 5

achieve a good trade-off between the overhead and deliveryratio For all the unknown messages carried by node 119894 whenthe destination of messages has higher similarity with node 119895

than it has with node 119894 node 119894 will replicate these messagesto node 119895 If the carrier similarity utility between node 119894

and node 119895 is greater than the predefined carrier similaritythreshold it means that node 119894 and node 119895 have similarmobility scope There is no need to increase the numberof message copies in the networks Then node 119894 will onlyforward these messages to node 119895 when the latter has a highdestination similarity utility

4 Buffer Management Scheme

Although multiple-copy routing can improve the probabilityof delivery rate it also inevitably brings more traffic intothe network When the node buffer overflows the networkperformance decreases sharply To overcome this problemwe propose a buffer management approach which exploitsmessages transmission status in the networks to decide thepriority of message replacement and scheduling

41 Message Transmission Analysis The dissemination ofmessages in the network is modeled from the perspective ofan individual message 119898 The nodes that hold message 119898 orits copy in the networks are called infected nodes other nodesare called uninfected nodes Let 119909 and 119904 denote the number ofinfected nodes and uninfected nodes respectively at a giventime Thus the total number of the network nodes can bewritten as

119873 = 119904+119909 (3)

Let 120582 represent the encounter rate between the nodes Eachinfected nodewillmeet120582119904119873 uninfected nodes in a unit timeThen the total number of infected nodes increases by 120582119904119909119873

in a unit time Let120583 denote the rate of droppingmessagesThenumber of nodes whose status is changed from infected touninfected is 120583119909 Then the increasing rate of infected nodesis

119889119909

119889119905=

120582119904119909

119873minus120583119909 (4)

Messages will be discarded when two encountered nodesexchange message copies and buffer overflow occurs Underthe condition that the encounter rate of nodes is greater thanthe dropping rate of messages that is 120582 gt 120583 let

120588 =120582

120583 (5)

Combining (3) (4) and (5) together yields

119889119909

119889119905= minus 120582119909 [

119909

119873minus(1minus

1120588)] (6)

The number of infected nodes can be expressed as

119909 = 119873(1minus1120588)

11 + [119873 (1 minus 1120588) minus 1] 119890minus120582(1minus1120588)119905

(7)

We can obtain from (5) and (7) that the increasing rate of 119909depends on its initial value Because the initial value of 119909 is1 119909 grows faster in the beginning stage of disseminationThelimit value increases as 120588 increases and is given by

119909 (infin) = 119873(1minus1120588) (8)

Let a utility function 119880(119909) model the delivery probabilityfor each single message 119880(119909) is an increasing function ofparameter 119909 In other words the greater the number of nodesreceiving the samemessage the higher delivery probability ofthe message This implies the following requirement on thederivative of the function 119880(119909)

119889119880 (119909)

119889119909ge 0 (9)

According to the law of diminishing marginal utility we canobtain the equation below

lim119909rarr119873

119889119880 (119909)

119889119909= 0 (10)

Equation (10) reflects the phenomenon that the improvementof the delivery probability is vanishing when a high deliveryprobability is reached that is a feasible assumption for generalcases

Note that in the case of buffer management the numberof infected nodes is not greater than the total number of nodesin the network Therefore there is an upper bound to theprobability for each single message For this reason we get

lim119909rarr119873

119880 (119909) = 1 (11)

It is often supposed that a utility function has some propertiesof regularity for example continuous differentiability at leastpiecewise When these properties are applied to (9) (10) and(11) we can have

exist119888

10158401015840

119880 (119909) lt 0 forall119909 ge 119888 (12)

Formula (12) implies that the concavity of119880(119909) at least for119909 is greater than a given valueTherefore when amessage hasfewer copies in the network119880(119909)has the properties of convexfunction From the perspective of the whole network if sucha message is dropped in the network the decreased utility ofthis message is larger than the increased utility when a nodeaccommodates a message with a large number of copies Onthe contrary when a message has a large number of copies inthe network 119880(119909) has the properties of concave function Ifsuch a message is dropped the decreased utility is less thanthe increased utility when a node accepts a message with fewcopies

From the above analysis we can obtain thatmessageswitha large number of copies in the network have much moretransfer opportunities These messages will have relativelyhigher probability to reach their destination nodes Whenwe control the increasing number of these messages thelost utility is limited at a low level For a single message it

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

4 International Journal of Distributed Sensor Networks

Carrier similarity is measured as the proportion of thesame history encountered nodes between themessage carriernode and the contact node If a node has high similarity withthe message carrier node it will have similar mobility scopeThe node will not be appropriate for diffusing message faraway The message carrier node will hold the messages byitself In contrast if a node has low similaritywith themessagecarrier node it will be capable of disseminating messages faraway Thus the node is a good potential relaying node Let 119878

119894

and 119878119895denote the set of encounters experienced by node 119894 and

node 119895 within 119879 respectively When node 119894 contacts node 119895the carrier similarity Sim119904

119894 calculated by node 119894 is defined as

follows

Sim119904119894=

10038161003816100381610038161003816119878119894cap 119878119895

100381610038161003816100381610038161003816100381610038161003816119878119894

1003816100381610038161003816

(1)

Destination similarity measures the number of the samenodes encountered by a node and the destination node If anode has high similarity with the destination it may have ahigher chance to meet the destination For a given node 119894 thedestination similarity is defined below

Sim119889119894=

1003816100381610038161003816119878119894 cap 119878119889

1003816100381610038161003816 (2)

where 119878119894and 119878

119889are the set of the encounter nodes expe-

rienced by node 119894 and the destination node within 119879respectively

33 Similarity-Based Routing Algorithm Here we proposethe similarity-based routing algorithm with the aim ofbringing a concise concept of behavior similarity into DTNrouting Similarity-based routing combines carrier and des-tination similarity to make routing decisions There are twointuitions behind this algorithm Carrier similarity couldcontrol the number of flooding message copies and avoidmissing the good relay node Destination similarity couldimprove forwarding efficiency

Each node maintains an encounter node vector (nv) Anentry in the nv is composed of two items including node IDand encounter timeThenode can obtain contact informationin regard to nodes that it has encountered and that it has notyet encountered When a node encounters other nodes thenode will exchange the nv If the encounter node already hasan entry in the nv the node will only modify the encountertime Otherwise a new entry will be inserted into the nvThenode will refresh its nv in a given time 119879 If the encountertime of a node exceeds the time 119879 the entry of the node willbe removed from the nv

Node contacts are represented as an 119899 times 119899 symmetricmatrix where 119899 is the number of contacts encountered by agiven node If there is a contact between node 119894 and node 119895the corresponding element in the matrix is set to 1 otherwiseto 0 The count of nonzero equivalent row entries in thematrix represents the number of common neighbors betweennode 119894 and node 119895 The intermediate nodes may obtain theuseful information in regard to the destination node viadirect encountered nodes The destination node may be anindirect node Then the similarity with indirect encounter

Upon meeting node jif hasMsgsForDest(j) == true thendeliverMsgs(j)

end ifexchangeEncountersVector(j nv)updateSimilarity( )exchangeSummaryVector( )if Sim119904

119894lt Similarity119904TH then

for all messages unknown to node j doif Sim119889

119895ge Sim119889

119894then

replicateMsgs(j messages)end if

end forelsefor all messages unknown to node j doif Sim119889

119895gt Sim119889

119894then

forwardMsgs(j messages)end if

end forend ifUpon reception messagem from node jupdateSummaryVector( )

Algorithm 1 Similarity-based routing algorithm SBR

nodes is also needed to be evaluated Node contacts withdirect encountered nodes and indirect encountered nodes arerepresented as an 119899 times 119896 matrix where 119899 is the number ofdirect encountered nodes and 119896 is the number of indirectencountered nodes More details of the calculation can befound in [11]

Each node also maintains a summary vector An entryin summary vector is composed of four items includingmessage ID replication number message TTL value andinitial message TTL value The summary vector is used toupdate the forwarding status of a message It is exchangedwhen two nodes are encountered When a node receives anew message a new entry is added in the summary vectorWhen the node discards a message the corresponding entryin the summary vector remains in the node until the TTLvalue of that message expires

Because the sizes of the encounter node vector and thesummary vector are very small compared to a message theoverheads of the encounter vector and the summary vectorare not considered

The routing algorithm is outlined inAlgorithm 1 It repre-sents the communication process whenmessage carrier node119894 meets the potential relaying node 119895 They exchange someinformation of their encounter history used for computingroutingmetrics like Epidemic Node 119894 calculates the similaritywith node 119895

If the carrier similarity utility between node 119894 andnode 119895 is below the predefined carrier similarity thresholdSimilarity119904TH it means that node 119895 has the ability to spreadthe message to more unfamiliar nodes If the threshold istoo small messages are hard to be sent to other nodes Ifthe threshold is too large excessive distribution of messagescannot be controlled Therefore Similarity119904TH is set as 05 to

International Journal of Distributed Sensor Networks 5

achieve a good trade-off between the overhead and deliveryratio For all the unknown messages carried by node 119894 whenthe destination of messages has higher similarity with node 119895

than it has with node 119894 node 119894 will replicate these messagesto node 119895 If the carrier similarity utility between node 119894

and node 119895 is greater than the predefined carrier similaritythreshold it means that node 119894 and node 119895 have similarmobility scope There is no need to increase the numberof message copies in the networks Then node 119894 will onlyforward these messages to node 119895 when the latter has a highdestination similarity utility

4 Buffer Management Scheme

Although multiple-copy routing can improve the probabilityof delivery rate it also inevitably brings more traffic intothe network When the node buffer overflows the networkperformance decreases sharply To overcome this problemwe propose a buffer management approach which exploitsmessages transmission status in the networks to decide thepriority of message replacement and scheduling

41 Message Transmission Analysis The dissemination ofmessages in the network is modeled from the perspective ofan individual message 119898 The nodes that hold message 119898 orits copy in the networks are called infected nodes other nodesare called uninfected nodes Let 119909 and 119904 denote the number ofinfected nodes and uninfected nodes respectively at a giventime Thus the total number of the network nodes can bewritten as

119873 = 119904+119909 (3)

Let 120582 represent the encounter rate between the nodes Eachinfected nodewillmeet120582119904119873 uninfected nodes in a unit timeThen the total number of infected nodes increases by 120582119904119909119873

in a unit time Let120583 denote the rate of droppingmessagesThenumber of nodes whose status is changed from infected touninfected is 120583119909 Then the increasing rate of infected nodesis

119889119909

119889119905=

120582119904119909

119873minus120583119909 (4)

Messages will be discarded when two encountered nodesexchange message copies and buffer overflow occurs Underthe condition that the encounter rate of nodes is greater thanthe dropping rate of messages that is 120582 gt 120583 let

120588 =120582

120583 (5)

Combining (3) (4) and (5) together yields

119889119909

119889119905= minus 120582119909 [

119909

119873minus(1minus

1120588)] (6)

The number of infected nodes can be expressed as

119909 = 119873(1minus1120588)

11 + [119873 (1 minus 1120588) minus 1] 119890minus120582(1minus1120588)119905

(7)

We can obtain from (5) and (7) that the increasing rate of 119909depends on its initial value Because the initial value of 119909 is1 119909 grows faster in the beginning stage of disseminationThelimit value increases as 120588 increases and is given by

119909 (infin) = 119873(1minus1120588) (8)

Let a utility function 119880(119909) model the delivery probabilityfor each single message 119880(119909) is an increasing function ofparameter 119909 In other words the greater the number of nodesreceiving the samemessage the higher delivery probability ofthe message This implies the following requirement on thederivative of the function 119880(119909)

119889119880 (119909)

119889119909ge 0 (9)

According to the law of diminishing marginal utility we canobtain the equation below

lim119909rarr119873

119889119880 (119909)

119889119909= 0 (10)

Equation (10) reflects the phenomenon that the improvementof the delivery probability is vanishing when a high deliveryprobability is reached that is a feasible assumption for generalcases

Note that in the case of buffer management the numberof infected nodes is not greater than the total number of nodesin the network Therefore there is an upper bound to theprobability for each single message For this reason we get

lim119909rarr119873

119880 (119909) = 1 (11)

It is often supposed that a utility function has some propertiesof regularity for example continuous differentiability at leastpiecewise When these properties are applied to (9) (10) and(11) we can have

exist119888

10158401015840

119880 (119909) lt 0 forall119909 ge 119888 (12)

Formula (12) implies that the concavity of119880(119909) at least for119909 is greater than a given valueTherefore when amessage hasfewer copies in the network119880(119909)has the properties of convexfunction From the perspective of the whole network if sucha message is dropped in the network the decreased utility ofthis message is larger than the increased utility when a nodeaccommodates a message with a large number of copies Onthe contrary when a message has a large number of copies inthe network 119880(119909) has the properties of concave function Ifsuch a message is dropped the decreased utility is less thanthe increased utility when a node accepts a message with fewcopies

From the above analysis we can obtain thatmessageswitha large number of copies in the network have much moretransfer opportunities These messages will have relativelyhigher probability to reach their destination nodes Whenwe control the increasing number of these messages thelost utility is limited at a low level For a single message it

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 5

achieve a good trade-off between the overhead and deliveryratio For all the unknown messages carried by node 119894 whenthe destination of messages has higher similarity with node 119895

than it has with node 119894 node 119894 will replicate these messagesto node 119895 If the carrier similarity utility between node 119894

and node 119895 is greater than the predefined carrier similaritythreshold it means that node 119894 and node 119895 have similarmobility scope There is no need to increase the numberof message copies in the networks Then node 119894 will onlyforward these messages to node 119895 when the latter has a highdestination similarity utility

4 Buffer Management Scheme

Although multiple-copy routing can improve the probabilityof delivery rate it also inevitably brings more traffic intothe network When the node buffer overflows the networkperformance decreases sharply To overcome this problemwe propose a buffer management approach which exploitsmessages transmission status in the networks to decide thepriority of message replacement and scheduling

41 Message Transmission Analysis The dissemination ofmessages in the network is modeled from the perspective ofan individual message 119898 The nodes that hold message 119898 orits copy in the networks are called infected nodes other nodesare called uninfected nodes Let 119909 and 119904 denote the number ofinfected nodes and uninfected nodes respectively at a giventime Thus the total number of the network nodes can bewritten as

119873 = 119904+119909 (3)

Let 120582 represent the encounter rate between the nodes Eachinfected nodewillmeet120582119904119873 uninfected nodes in a unit timeThen the total number of infected nodes increases by 120582119904119909119873

in a unit time Let120583 denote the rate of droppingmessagesThenumber of nodes whose status is changed from infected touninfected is 120583119909 Then the increasing rate of infected nodesis

119889119909

119889119905=

120582119904119909

119873minus120583119909 (4)

Messages will be discarded when two encountered nodesexchange message copies and buffer overflow occurs Underthe condition that the encounter rate of nodes is greater thanthe dropping rate of messages that is 120582 gt 120583 let

120588 =120582

120583 (5)

Combining (3) (4) and (5) together yields

119889119909

119889119905= minus 120582119909 [

119909

119873minus(1minus

1120588)] (6)

The number of infected nodes can be expressed as

119909 = 119873(1minus1120588)

11 + [119873 (1 minus 1120588) minus 1] 119890minus120582(1minus1120588)119905

(7)

We can obtain from (5) and (7) that the increasing rate of 119909depends on its initial value Because the initial value of 119909 is1 119909 grows faster in the beginning stage of disseminationThelimit value increases as 120588 increases and is given by

119909 (infin) = 119873(1minus1120588) (8)

Let a utility function 119880(119909) model the delivery probabilityfor each single message 119880(119909) is an increasing function ofparameter 119909 In other words the greater the number of nodesreceiving the samemessage the higher delivery probability ofthe message This implies the following requirement on thederivative of the function 119880(119909)

119889119880 (119909)

119889119909ge 0 (9)

According to the law of diminishing marginal utility we canobtain the equation below

lim119909rarr119873

119889119880 (119909)

119889119909= 0 (10)

Equation (10) reflects the phenomenon that the improvementof the delivery probability is vanishing when a high deliveryprobability is reached that is a feasible assumption for generalcases

Note that in the case of buffer management the numberof infected nodes is not greater than the total number of nodesin the network Therefore there is an upper bound to theprobability for each single message For this reason we get

lim119909rarr119873

119880 (119909) = 1 (11)

It is often supposed that a utility function has some propertiesof regularity for example continuous differentiability at leastpiecewise When these properties are applied to (9) (10) and(11) we can have

exist119888

10158401015840

119880 (119909) lt 0 forall119909 ge 119888 (12)

Formula (12) implies that the concavity of119880(119909) at least for119909 is greater than a given valueTherefore when amessage hasfewer copies in the network119880(119909)has the properties of convexfunction From the perspective of the whole network if sucha message is dropped in the network the decreased utility ofthis message is larger than the increased utility when a nodeaccommodates a message with a large number of copies Onthe contrary when a message has a large number of copies inthe network 119880(119909) has the properties of concave function Ifsuch a message is dropped the decreased utility is less thanthe increased utility when a node accepts a message with fewcopies

From the above analysis we can obtain thatmessageswitha large number of copies in the network have much moretransfer opportunities These messages will have relativelyhigher probability to reach their destination nodes Whenwe control the increasing number of these messages thelost utility is limited at a low level For a single message it

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

6 International Journal of Distributed Sensor Networks

decreases their delivery probability but brings more transferopportunities for messages that have fewer copies in the net-work The increased utility of these messages is greater thanthe decreased utility of messages that have more copies in thenetworksTherefore it will improve the overall performance

42 Buffer Replacement Scheme Because of intermittentconnectivity in the network a node could not get the accurateglobal status about a particular message It can use statisticallearning to estimate the dissemination status of a messagewhen nodes are encountered We introduce two metrics tomeasure the priority of a message including the number ofmessage copies and the dissemination speed of a message Amessage that has a smaller replication number is assigned ahigher priority Ifmessages have the same replication numberthemessage with the lower speed of dissemination is assigneda higher priority

Let 119877119894

119898denote the replication number of message 119898

known by node 119894 Obviously it can be seen that message 119898which has the greater value of 119877119894

119898 has the strong ability to

spread in the network Meanwhile more copies of message119898 might reside in the network On the contrary message 119898

that has a lower value of 119877119894119898might leave fewer copies in the

network Node 119894 will discard the message that has a highervalue of 119877119894

119898first

Here we describe the process of learning the replicationnumber of a single message The initial value of replicationnumber is set to 1 when a new message is generated in thenetwork When node 119894 that carries message 119898 meets node119895 that does not carry message 119898 the replication number ofmessage 119898 is processed in the following two cases

(i) One case is that node 119895 is selected as a relay nodefor message 119898 If node 119895 does not contain anyinformation aboutmessage119898 the replication numberofmessage119898 is set to119877

119894

119898+1 in both nodes Otherwise

both nodes exchange the summary vector and set thevalue to max(119877119894

119898 119877119895

119898) + 1

(ii) The other case is that node 119895 is not selected as a relaynode for message 119898 When node 119895 does not containany information about message 119898 the replicationnumber of message 119898 is set to 119877

119894

119898in both nodes

Otherwise both nodes exchange the summary vectorand set the value to max(119877119894

119898 119877119895

119898)

When different messages have the same estimated repli-cation number we use theRatemetric to describe the dissem-ination speed of a message According to the path explosionphenomenon once a message reaches the destination thereare a number of near-optimal paths to the destinationTherefore more message copies can exist in the networksPath explosion occurs much faster among the higher contactrate nodes than the lower contact rate nodes The Ratemetric can reflect the nodes contact rate from viewpoint ofa message This metric is defined as follows

Rate =119870119898

TTLinit minus TTL (13)

Upon receiving messagem from the encounterednodewhile BufferfreeSize lt 119898size domsg = minPriority(messages in Buffer ⋃119898)if msg == 119898 thendeleteMessage(msg)

elsedeleteMessage(msg)BufferfreeSize += msgsize

end ifend while

Algorithm 2 Buffer replacement algorithm MTSBR

where 119870119898represents the hop count experienced by message

119898 Our design for the buffermanagement scheme associates ahop count with each messageThe hop count119870

119898is estimated

according to the message replication number The originalreplication number is one For a given message 119898 carried bynode 119894 the hop count 119870

119898= 119877119894

119898minus 1 The message that has a

higher dissemination speed might have many more copies inthe networks

The node will accept a new message if it has enoughfree buffer space Otherwise the node will compare all themessages in its buffer with the new one according to thepriority discussed aboveMessagewith a lower prioritywill bediscarded The algorithmMTSBR is shown in Algorithm 2

43 Buffer Scheduling Scheme A set of messages that aredetermined by routing protocol should be forwarded to abetter intermediate node We call these messages Ready Set(RS) Ideally message carrier node will transmit all of them tothe relay node Unfortunately not all the messages could betransmitted due to finite bandwidth or unexpected interrup-tions It is important for a node to decide the order in whichthe messages are transmitted Meanwhile routing protocoldoes not consider whether the relay node has enough bufferspace to hold thesemessages Obviously bandwidth and nodeenergy are wasted when transmitted messages are droppeddue to buffer overflow It is also important for a node todecide which messages should be forwarded to relay nodeTo address these problems we propose MTSBS schedulingscheme that is outlined in Algorithm 3

Firstly MTSBS sorts messages in RS in a descendingorder according to their priority In all cases the messagewith higher priority will be forwarded Secondly MTSBS willchoose whichmessages to forward to relay node If the lowestpriority of messages in RS is greater than the highest priorityof messages in the peering node then the node forwards allthe messages If the highest priority of messages is lower thanthe lowest priority in the peering node the node will onlyforward messages that the peering node could contain in itsfree buffer In other cases node merges the message list in RSand peering node Then it sorts the merged list and selectsthe top messages that their buffer occupancy is close to thebuffer capacity These messages residing in the local node areforwarded to relay node

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 7

119872119894 a set of sorted messages that selected by routing algorithm in node 119894

119872119894(119896) the (119896 + 1)th message in 119872

119894

if 119872119894occupancy lt node 119895rsquos freebuffer then

Sending119872119894to node 119895

end ifif the lowest priority in 119872

119894gt the highest priority in119872

119895

thenSending119872

119894to node 119895

end ifif the highest priority in119872

119894lt the lowest priority in 119872

119895

thenfor 119896 = 0 119896 lt 119872

119894size( ) 119896++ do

if 119872occupancy lt node 119895rsquos freebuffer then119872add(119872

119894(119896))

end ifend forSending119872 to node 119895

end ifSending (TopbuffSize(119872119894 + 119872

119895) minus 119872

119895) to node 119895

Algorithm 3 Buffer scheduling algorithm MTSBS

5 Performance Evaluation

We compare the performance of the proposed SBR algorithmagainst the following three routing algorithms (Epidemic [13]Prophet [21] and ProphetV2 [22]) in DTNs using the ONE[39] simulator

Epidemic [13] Messages are flooded to all the encounterednodes It uses the DO (Drop Oldest message that has theshortest TTL value is dropped first) and adopts random strat-egy for message replacement and scheduling respectively Itis the benchmark that was used for performance analysis andcomparison in the previous works

Prophet [21] This is a mobility-based approach in DTNs Itcalculates the routing metric by using the history of nodeencounters and transitivity A message is forwarded to anode that has a higher estimated delivery predictability fora specific destination node than the current message carriernode It also uses the DO replacement strategy and adoptsGRTRMax for message scheduling GRTRMax forwardsmessages in descending order of delivery predictabilities

ProphetV2 [22] It redefines the transitivity update equationand direct encounter update equation in Prophet

In this experiment we also evaluate SBR with differentbuffer management schemes SBR-1 denotes SBR routingalgorithm with DO replacement and random schedulingscheme SBR-2 represents SBR algorithm with HBD (His-tory Based Drop) [32] replacement and random schedulingscheme HBD is a distributed message replacement schemebased on the estimated global information about messages tooptimize the specific metric SBR-3 represents SBR algorithmwith our proposed buffer management scheme

We compare the performance of these algorithms interms of message delivery ratio overhead ratio and averagedelay

100m

Figure 1 Map-based scenario

Delivery ratio is defined as the ratio of the number ofdelivered messages to the total number of sent messages

Overhead ratio is the average number of relays used forone deliveredmessage As the size of a summary vector is verysmall compared to a message the overhead of the summaryvector is not considered

Average delay refers to the mean of time from messagesgeneration to their copies first received by the destinationnodes

51 Experimental Settings Two mobility models that is amap-based mobility model and RWP mobility model areused to evaluate the performance of routing protocols

Under the map-based model we use the default map inONE which consists of a 4500m times 3500m area The map-based scenario is shown in Figure 1 Each labeled circle in themap represents the node which belongs to a specific group Inorder to investigate the impact of different number of groupswe compare these routing algorithms with 3 and 4 groupsrespectively We set 119896 (119896 = 3 4) Points-of-Interest (POIs)

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

8 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

2800

2400

2000

1600

1200

800

400

0

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

400

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 2 Map-based (3 groups) influence of buffer size on performance

which belong to a certain POI group Each node movesamong the POI groups with a specific probability Nodesmove to other POI groups with the probability Pr (Pr =

01) Nodes move in the local group with the probability1minus(119896minus1)PrThe setting of destination selection probability issimilar to the one that was done in [21] Nodes in each grouprepresent pedestrians These nodes move with the averagespeed of 134ms which represents a pedestrian averagewalking speed [40]

Under the RWPmodel the simulation area is 1 kmtimes 1 kmNodes are randomly distributed in the field Nodes have anaveragemoving speed of 134ms and the pause time of a stopis uniformly distributed in [0 120] seconds

For the two simulation models each node uses an idealcommunication module and has a communication range of10m The transmission speed of nodes is 2Mbps Simulationtime is 4 hours to ensure that the nodes can form the steadymobility pattern and the stable simulation results can beachieved A new message with TTL is generated every 15seconds The size of messages is 1 KB

52 Experimental Results

Varying Buffer Size Figures 2 and 3 reveal the impact of buffersize on the performance of routing algorithms in the 3- and4-group conditions under the map-based mobility model

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 9

09

08

07

06

05

04

03

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3500

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 3 Map-based (4 groups) influence of buffer size on performance

respectively In the two scenarios pedestrian groups have 40nodes in each group Their buffer size varies from 100 to600KB As the results in the 3- and 4-group conditions showa similar trend we only discuss the results for the 3 groupsrsquocondition

Figure 2(a) shows that the delivery ratio becomes largeras the buffer size increases Epidemic floods more copiesin the networks so it has the lowest delivery ratio whenthe node buffer space is very small As expected ProphetV2significantly outperforms Prophet in the map-based mobilitymodel because ProphetV2 can deal with the problem thatnodes come together and repeatedly exchange their setsof delivery predictabilities SBR limits the flooding and

improves the delivery probability As for the buffer man-agement schemes MTSBR could guarantee the transmissionefficiency as it incorporates network status to make decisionWhen the buffer overflowsMTSBR dropsmessages that havethe most copies Although HBD and DO take the numberof message copies into consideration they do not care aboutdissemination capacity of messages Since DO considers onlythe number of message copies in a local view and does notincorporate network status SBR-3 has a higher delivery ratiocompared to SBR-1 and SBR-2

It can be seen from Figure 2(b) that three SBR algorithmshave the lower overhead ratio than Epidemic Prophet andProphetV2 Epidemic replicates message to any encountered

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

10 International Journal of Distributed Sensor Networks

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

00

(a) Delivery ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Ove

rhea

d ra

tio

200

160

120

80

40

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

2200

2000

1800

1600

(c) Average delay

Figure 4 RWP influence of buffer size on performance

nodes When the node buffer space is very small moremessage copies are discarded and retransmitted So it causesthe higher overhead ratio Prophet and ProphetV2 onlyreplicate messages to the encountered nodes that have ahigher delivery probability SBR algorithms can alleviatetraffic to some extent because they are able to control thenumber ofmessage copies by comparing the carrier similaritybetween encountered nodes SBR-3 has the lowest overheadratio among all the SBR algorithmsTheMTSBR replacementscheme could reduce the number of retransmissions It couldpartially avoid dropping messages that is in the beginningstage of dissemination MTSBS will decide which messages

to transmitThe scheduler considers the buffer constraint andwill not transmit themessages that will be dropped in the nextintermediate node So it has a relatively low overhead ratio

Figure 2(c) shows that the average delay of all the routingalgorithms decreases When the buffer size increases moremessage copies will be saved in the nodesrsquo bufferThemessagecopies will have more opportunities to arrive at the destina-tion node Therefore the message delay will decrease SBRalgorithms performance in terms of average delay remainsacceptable especially SBR-3 algorithm

Figure 4 shows the impact of buffer size on the perfor-mance of routing algorithms under the RWPmobility model

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 11

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

2400

2000

1600

1200

800

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

(c) Average delay

Figure 5 Map-based (3 groups) influence of the number of nodes on performance

Epidemic gains a significant benefit from increased buffersize It has the lowest average delay when the buffer sizeexceeds 200KB in this experiment SBR algorithms still havethe high delivery ratio when buffer space is small especiallySBR-3 algorithm These results show that our approach canchoose appropriate relay nodes with lower overhead Theaverage delay is also acceptable The RWP mobility modeldoes not provide predictable mobility patterns that Prophetand ProphetV2 can leverage So the difference between theirperformances is not great like that in group conditions

Varying Number of Nodes Figures 5 6 and 7 depict theimpact of the increasing number of nodes on the performance

of different protocols The number of nodes in each groupvaries from 20 to 50 under the map-based mobility modelUnder the RWP mobility model the number of nodes variesfrom 60 to 150

Figure 5(a) reveals that the delivery ratio does not fluc-tuate much when the number of nodes increases in eachgroup For the heavy traffic buffer contention will becomemore serious when the number of nodes increases even if therouting protocols adopt the controlled flooding scheme SinceSBR uses message transmission status to manage the bufferspace it has a higher delivery ratio When the number of net-work nodes increases the estimate of the global informationin HBD becomes more difficult It has less impact on routing

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

12 International Journal of Distributed Sensor Networks

09

08

07

06

05

04

Del

iver

y ra

tio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

(a) Delivery ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

3000

2500

2000

1500

1000

500

0

Ove

rhea

d ra

tio(b) Overhead ratio

20 30 40 50

Number of nodes per groupProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

2500

2000

1500

1000

Aver

age d

elay

(s)

(c) Average delay

Figure 6 Map-based (4 groups) influence of the number of nodes on performance

thanMTSBRThus SBR-3 has higher delivery ratio than SBR-2

Figure 5(b) shows that as the number of network nodesincreases the transmission opportunities are also increasedAs more message copies are sent in the network the buffercontention becomes more serious thus resulting in theincreasing overhead ratio The effect of buffer managementscheme for SBR becomes obvious MTSBR limits furtherspread of messages that might have a large number of copiesand MTSBS could reduce unnecessary transmissions causedby buffer overflow Therefore the increase of overhead ratioin SBR-3 is low

It can be seen fromFigure 5(c) that the average delay of allthe routing algorithms decreases Because more nodes par-ticipate in the relay activity more forwarding opportunitiesarise It can alleviate the impact of forwarding limitation ofSBR algorithms Prophet and ProphetV2 It is worth notingthat Figure 6 reveals the performance trends similar to thosein Figure 5

Figure 7 shows the results under the RWP mobilitymodel Communication opportunities arise when the nodedensity increases We can see from Figure 7(a) that SBRalgorithms have relatively higher delivery ratio than otheralgorithms when the number of nodes increases Figure 7(b)

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 13

Del

iver

y ra

tio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

08

06

04

02

10

(a) Delivery ratio

60 90 120 150

Number of nodes

ProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

0

Ove

rhea

d ra

tio

160

120

80

40

(b) Overhead ratio

60 90 120 150

Number of nodesProphetV2

ProphetEpidemic

SBR-1SBR-2SBR-3

Aver

age d

elay

(s)

2400

3600

3000

1800

1200

600

(c) Average delay

Figure 7 RWP influence of the number of nodes on performance

shows that the overhead increases when the number of nodesincreases But SBR algorithms can control the number of dis-seminated messages by carrier similarity to some extent Ourproposed buffer management scheme can further alleviatemessage retransmission so SBR-3 has the lowest overheadratio In Figure 7(c) we can see that SBR algorithms also havegood performance in terms of average delay

Similarity Threshold Analysis Here to analyze the impact ofsimilarity threshold Similarity119904TH on protocol performancewe evaluate SBR-1 under different Similarity119904TH value with

025 05 and 075 respectively The buffer size varies from100 to 600KB under the map-based mobility model As canbe seen from Figure 8 the SBR-1 has similar trends for theperformancewith Similarity119904TH varying In the Similarity119904TH =

025 case SBR-1 has the lowest overhead ratio becausemessages are difficult to be sent to other nodes HoweverSBR-1 has the lowest delivery ratio and the highest averagedelay In the Similarity119904TH = 075 case SBR-1 has goodperformance on delivery ratio and average delay But it alsobrings more overhead For Similarity119904TH = 05 SBR-1 has thehighest delivery ratio The overhead ratio and average delay

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

14 International Journal of Distributed Sensor Networks

09

08

07

06

05

Del

iver

y ra

tio

100 200 300 400 500 600

Buffer size (KB)

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(a) Delivery ratio

100

200

300

400

500

600

700

100 200 300 400 500 600

Buffer size (KB)

Ove

rhea

d ra

tio

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(b) Overhead ratio

100 200 300 400 500 600

Buffer size (KB)

Aver

age d

elay

(s)

2400

2000

1200

1600

SimilaritysTH = 025

SimilaritysTH = 05

SimilaritysTH = 075

(c) Average delay

Figure 8 Comparison of performance under different similarity threshold

are on themiddle level It achieves a good performance trade-off among delivery ratio overhead ratio and average delay

6 Conclusions

The existing routing protocols in DTNs have considered onlythe similarity of mobility patterns between the relaying nodesand the destination nodes In this paper we take into accountthe similarity of mobility patterns between the message car-rier node and its encountered nodes and propose a similarity-based routing protocol which uses different similarity as

the condition of replication or forwarding Moreover weestimate the replication number and spreading speed ofmessages using encounter historyThemessagewith a smallerreplication number and lower speed of dissemination isassigned the higher priority Furthermore we propose abuffer replacement schemeMTSBR and a scheduling schemeMTSBS according to the priority Simulation results showthat our routing protocols combined with the buffer man-agement schemes outperform the existing routing protocolsin terms of delivery ratio and overhead ratio in guaranteeddelay

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 15

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This project is supported by the National Natural Sci-ence Foundation of China (Grants nos 61103204 61273232and 61402541) the Humanities and Social Science YouthFoundation of Ministry of Education of China (Grant no13YJCZH110) the Construct Program of the Key Disciplinein Hunan Province the Scientific Research Fund of HunanProvincial Education Department (Grant no 12C0768) theMajor Science and Technology Research Program for Strate-gic Emerging Industry of Hunan (Grant no 2012GK4054)the Open Funding of Science and Technology on Informa-tion Transmission and Dissemination in CommunicationNetworks Laboratory (Grant no ITDU14010KX142600017)and Postdoctoral Funding of Central South University andChangsha Bohua Technology Co Ltd China

References

[1] K Fall ldquoA delay-tolerant network architecture for challengedinternetsrdquo in Proceedings of the Conference on ApplicationsTechnologies Architectures and Protocols for Computer Com-munications (SIGCOMM rsquo03) pp 27ndash34 ACM KarlsruheGermany August 2003

[2] K K Sevimli and M Soyturk ldquoEnabling delay-tolerant com-munications for partially connected vehicular ad hoc networksrdquoInternational Journal of Ad Hoc and Ubiquitous Computing vol11 no 2-3 pp 157ndash168 2012

[3] S Ehsan K Bradford M Brugger et al ldquoDesign and analysisof delay-tolerant sensor networks for monitoring and trackingfree-roaming animalsrdquo IEEE Transactions on Wireless Commu-nications vol 11 no 3 pp 1220ndash1227 2012

[4] P Hui A Chaintreau J Scott R Gass J Crowcroft andCDiotldquoPocket switched networks and human mobility in conferenceenvironmentsrdquo in Proceedings of the ACM SIGCOMM Work-shop on Delay-Tolerant Networking (WDTN rsquo05) pp 244ndash251ACM Philadelphia Pa USA August 2005

[5] Y Xie and GWang ldquoMessage matching-based greedy behaviordetection in delay tolerant networksrdquo Journal of Computer andSystem Sciences vol 80 no 5 pp 903ndash915 2014

[6] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks thesingle-copy caserdquo IEEEACM Transactions on Networking vol16 no 1 pp 63ndash76 2008

[7] T Spyropoulos K Psounis and C S Raghavendra ldquoEffi-cient routing in intermittently connected mobile networks themultiple-copy caserdquo IEEEACM Transactions on Networkingvol 16 no 1 pp 77ndash90 2008

[8] Y Liu J Wang S Zhang and H Zhou ldquoA buffer managementscheme based on message transmission status in delay tolerantnetworksrdquo in Proceedings of the IEEE Global Telecommunica-tions Conference (GLOBECOM rsquo11) pp 1ndash5 IEEE Houston TexUSA December 2011

[9] Q Yuan I Cardei and J Wu ldquoAn efficient prediction-basedrouting in disruption-tolerant networksrdquo IEEE Transactions onParallel and Distributed Systems vol 23 no 1 pp 19ndash31 2012

[10] J Leguay T Friedman and V Conan ldquoDTN routing in amobility pattern spacerdquo in Proceedings of the ACM SIGCOMMWorkshop on Delay-Tolerant Networking (WDTN rsquo05) pp 276ndash283 ACM Philadelphia Pa USA August 2005

[11] EMDaly andMHaahr ldquoSocial network analysis for routing indisconnected delay-tolerantMANETsrdquo in Proceedings of the 8thACM International Symposium on Mobile Ad Hoc Networkingand Computing (MobiHoc rsquo07) pp 32ndash40 September 2007

[12] Q Li W Gao S Zhu and G Cao ldquoA routing protocol forsocially selfish delay tolerant networksrdquo Ad Hoc Networks vol10 no 8 pp 1619ndash1632 2012

[13] A Vahdat and D Becker ldquoEpidemic routing for partiallyconnected ad hoc networksrdquo Tech Rep CS-200006 DukeUniversity 2000

[14] Y Wang S Jain M Martonosi and K Fall ldquoErasure-codingbased routing for opportunistic networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 229ndash236 ACM Philadelphia Pa USAAugust2005

[15] F Tsapeli and V Tsaoussidis ldquoRouting for opportunistic net-works based on probabilistic erasure codingrdquo inWiredWirelessInternet Communication 10th International Conference WWIC2012 Santorini Greece June 6ndash8 2012 Proceedings vol 7277of Lecture Notes in Computer Science pp 257ndash268 SpringerBerlin Germany 2012

[16] J Widmer and J-Y Le Boudec ldquoNetwork coding for effi-cient communication in extreme networksrdquo in Proceedings ofthe ACM SIGCOMM Workshop on Delay-Tolerant Networking(WDTN rsquo05) pp 284ndash291 August 2005

[17] E Altman L Sassatelli and F D Pellegrini ldquoDynamic controlof coding for progressive packet arrivals in DTNsrdquo IEEETransactions onWireless Communications vol 12 no 2 pp 725ndash735 2013

[18] T Spyropoulos K Psounis and C S Raghavendra ldquoSpray andwait an efficient routing scheme for intermittently connectedmobile networksrdquo in Proceedings of the ACM Workshop onDelay-Tolerant Networking (SIGCOMM rsquo05) pp 252ndash259 2005

[19] S C Nelson M Bakht and R Kravets ldquoEncounter-based rout-ing inDTNsrdquo in Proceedings of the 28th Conference on ComputerCommunications (INFOCOM rsquo09) pp 846ndash854 IEEE April2009

[20] A Elwhishi P-H Ho K S Naik and B Shihada ldquoSelf-adaptivecontention aware routing protocol for intermittently connectedmobile networksrdquo IEEETransactions on Parallel andDistributedSystems vol 24 no 7 pp 1422ndash1435 2013

[21] A Lindgren A Doria and O Schelen ldquoProbabilistic routing inintermittently connected networksrdquo ACM SIGMOBILE MobileComputing and Communications Review vol 7 no 3 pp 19ndash202003

[22] S Grasic E Davies A Lindgren and A Doria ldquoThe evolutionof a DTN routing protocolmdashPRoPHETv2rdquo in Proceedings of the6th ACMWorkshop on Challenged Networks (CHANTS rsquo11) pp27ndash30 ACM Las Vegas Nev USA September 2011

[23] R Ramanathan R Hansen P Basu R Rosales-Hain andR Krishnan ldquoPrioritized epidemic routing for opportunisticnetworksrdquo in Proceedings of the 5th International Conference onMobile Systems Applications and Services (MobiOpprsquo 07) pp62ndash66 June 2007

[24] A Balasubramanian B N Levine and A VenkataramanildquoReplication routing in DTNs a resource allocation approachrdquoIEEEACM Transactions on Networking vol 18 no 2 pp 596ndash609 2010

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

16 International Journal of Distributed Sensor Networks

[25] J Huang J Wang and J Ye ldquoA buffer management algorithmfor improving updown transmission congestion protocol fair-ness in IEEE 80211 wireless local area networksrdquo InternationalJournal of Communication Systems vol 27 no 10 pp 2228ndash2240 2014

[26] J Wang L Rong and Y Liu ldquoA robust proportional controllerfor AQM based on optimized second-order system modelrdquoComputer Communications vol 31 no 10 pp 2468ndash2477 2008

[27] J Wang L Rong and Y Liu ldquoDesign of a stabilizing AQMcontroller for large-delay networks based on internal modelcontrolrdquo Computer Communications vol 31 no 10 pp 1911ndash1918 2008

[28] J Wang P Dong J Chen J Huang S Zhang and W WangldquoAdaptive explicit congestion control based on bandwidth esti-mation for high bandwidth-delay product networksrdquoComputerCommunications vol 36 no 10-11 pp 1235ndash1244 2013

[29] T Zhang J Wang J Huang Y Huang J Chen and Y PanldquoAdaptive-acceleration data center TCPrdquo IEEE Transactions onComputers vol 64 no 6 pp 1522ndash1533 2014

[30] J Ye J Huang J Wang S Zhang and Z Zhang ldquoECN-basedcongestion probability prediction over hybrid wired-wirelessnetworksrdquo International Journal of Distributed Sensor Networksvol 2014 Article ID 134620 11 pages 2014

[31] X Zhang G Neglia J Kurose and D Towsley ldquoPerformancemodeling of epidemic routingrdquo Computer Networks vol 51 no10 pp 2867ndash2891 2007

[32] A Krifa C Barakat and T Spyropoulos ldquoOptimal buffer man-agement policies for delay tolerant networksrdquo in Proceedingsof the 5th Annual IEEE Communications Society Conferenceon Sensor Mesh and Ad Hoc Communications and Networks(SECON rsquo08) pp 260ndash268 San Francisco Calif USA June2008

[33] A Lindgren and K S Phanse ldquoEvaluation of queueing poli-cies and forwarding strategies for routing in intermittentlyconnected networksrdquo in Proceedings of the 1st InternationalConference on Communication System Software andMiddleware(COMSWARE rsquo06) pp 1ndash10 IEEE New Delhi India January2006

[34] V Erramilli and M Crovella ldquoForwarding in opportunisticnetworks with resource constraintsrdquo in Proceedings of the 3rdACMWorkshop on Challenged Networks (CHANTS rsquo08) pp 41ndash48 ACM San Francisco Calif USA September 2008

[35] V Erramilli M Crovella A Chaintreau and C Diot ldquoDelega-tion forwardingrdquo in Proceedings of the 9th ACM InternationalSymposium on Mobile Ad Hoc Networking and Computing(MobiHoc rsquo08) pp 251ndash260 May 2008

[36] S Rashid A H Abdullah M S M Zahid and Q Ayub ldquoMeandrop an effectural buffer management policy for delay tolerantnetworkrdquo European Journal of Scientific Research vol 70 no 3pp 396ndash407 2012

[37] T Spyropoulos K Psounis and C S Raghavendra ldquoPerfor-mance analysis of mobility-assisted routingrdquo in Proceedingsof the 7th ACM International Symposium on Mobile Ad HocNetworking and Computing (MobiHoc rsquo06) pp 49ndash60 May2006

[38] R Groenevelt P Nain and G Koole ldquoMessage delay inMANETrdquo ACM SIGMETRICS Performance Evaluation Reviewvol 33 no 1 pp 412ndash413 2005

[39] A Keranen J Ott and T Karkkainen ldquoThe ONE simulator forDTN protocol evaluationrdquo in Proceedings of the 2nd Interna-tional Conference on Simulation Tools and Techniques (Simutoolsrsquo09) pp 1ndash10 ACM Rome Italy March 2009

[40] J Ghosh S J Philip and C Qiao ldquoSociological orbit awarelocation approximation and routing (SOLAR) in MANETrdquo AdHoc Networks vol 5 no 2 pp 189ndash209 2007

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of