anenergy-awaretaskoffloadingmechanisminmultiuser mobile...

Research ArticleAn Energy-Aware Task Offloading Mechanism in MultiuserMobile-Edge Cloud Computing

Lan Li Xiaoyong Zhang Kaiyang Liu Fu Jiang and Jun Peng

School of Information Science and Engineering China Hunan Engineering Laboratory of Rail Vehicles Braking TechnologyCentral South University Changsha Hunan 410083 China

Correspondence should be addressed to Xiaoyong Zhang zhangxycsueducn

Received 31 March 2017 Revised 9 November 2017 Accepted 31 December 2017 Published 2 April 2018

Academic Editor Bartolomeo Montrucchio

Copyright copy 2018 Lan Li et al -is 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

Mobile-edge cloud computing an emerging and prospective computing paradigm can facilitate the complex application ex-ecution on resource-constrained mobile devices by offloading computation-intensive tasks to the mobile-edge cloud server whichis usually deployed in close proximity to the wireless access point However in the multichannel wireless interference envi-ronment the competition of mobile users for communication resources is not conducive to the energy efficiency of task off-loading-erefore how to make the offloading decision for each mobile user and select its suitable channel become critical issuesIn this paper the problem of the offloading decision is formulated as a 0-1 nonlinear integer programming problem under theconstraints of channel interference threshold and the time deadline -rough the classification and priority determination for themobile devices a reverse auction-based offloading method is proposed to solve this optimization problem for energy efficiencyimprovement -e proposed algorithm not only achieves the task offloading decision but also gives the facility of resourceallocation In the energy efficiency performance aspects simulation results show the superiority of the proposed scheme

1 Introduction

Internet of things (IoT) devices such as sensors andwearable devices are increasingly penetrating into our ev-eryday lives Gartner forecasted that by 2020 the Internet ofthings (IoT) devices will reach 50 billion representing analmost 30-fold increase from 09 billion in 2009 [1] How-ever it is well known that mobile devices have their inherentproblems such as finite computing power and particularlylimited battery life [2] -ese resource-limited mobile de-vices are difficult to support computation-intensive appli-cations such as interactive gaming imagevideo processingand online social networking services [3] -erefore whendealing with the sophisticated applications on devices thecontradiction between the requirements of network appli-cations and the limited resource of mobile devices posesa significant challenge

A new architecture and technology known as MobileCloud Computing (MCC) brings a new idea which canaugment the processing ability of mobile devices andreduce the energy consumption of mobile devices at the

meantime by migrating computational tasks from mobiledevices to infrastructure-based cloud servers [4 5] -usMCC has the potential to address the aforementionedchallenge [6]

In recent years cloud offloading technologies have beenwidely studied by researchers all over the world Consideringthe huge increase of computational demand on the mobiledevices in the 5G networking Chen [7] proposed a gametheoretic approach for the distributed task offloading de-cision to improve the energy efficiency Barbarossa et al [8]investigated the problem of multiuser computation off-loading in multiradio channel scenarios ignoring thecommunication interference In order to save the energyconsumption of mobile devices and meet the requirementsof the application execution time Huang et al [9] proposeda Lyapunov optimization-based algorithm for dynamicoffloading which can reduce the computational complexityat the same time to obtain a near-optimal offloading de-cision In large-scale mobile applications Yang et al [10]researched the problem of multiuser computing partition forthe purpose of minimizing the average task completion of all

HindawiMobile Information SystemsVolume 2018 Article ID 7646705 12 pageshttpsdoiorg10115520187646705

users and designed an offline heuristic offloading algorithmViswanathan et al [11] proposed a resource provisioningframework for organizing the heterogeneous devices in thevicinity A joint optimization framework of the wirelessresource and computing resource is proposed in [12] for theenergy-constrained mobile users in the femtocell

In order to solve the problem of transmission delay andenergy consumption Satyanarayanan et al [13] proposed anarchitecture replacing the remote cloud with nearby cloud-lets Zhang et al [14] proposed a Markov decision processoffloading algorithm for mobile users in an intermittentlyconnected cloudlet system However the cloudlet-basedmobile cloud computing has some drawbacks Due to thelimited wireless network coverage the cloudlet cannotguaranteeubiquitous service everywhere forusersMoreoverthe computation resource of the cloudlet is insufficient tosatisfy theQoS requirements of a large number of users in thefuture

-ese existing task offloading strategies in mobile cloudcomputing are not sufficient to greatly improve the energyefficiency of the system-e long transmission delay betweenthe mobile device and traditional cloud servers is a criticalissue which is the inherent limitation of mobile cloudcomputing [15 16] -e long propagation distance from themobile device to the remote cloud data center will result inunacceptable long latency for mobile applications -e ad-ditional communication transmission delay will decrease thecomputation offloading efficiency and the QoS of users

-erefore a mobile-edge cloud (MEC) computing ar-chitecture is adopted With the development of wirelesscommunication technologies such as Wi-Fi 4G and 5G theMEC is envisioned as a promising and challenging approachto address the abovementioned challenges [17] In the MECframework mobile devices are able to offload their tasks tothe MEC clouds through the radio access points nearbyrather than the public clouds such as Amazon EC2 andWindows Azure -us this MEC paradigm can providelower latency and high communication rate and computingagility in the process of computation offloading

-ere are a few studies on the efficient computationoffloading mechanism of the MEC For instance Wang et al[18] used the Markov decision process to formulate a se-quential offloading decision-making problem for dynamicservice migration Considering the finite number of wirelessaccess channels and the interference Chen et al [19] pre-sented a distributed offloading decision method based ongame theory Sardellitti et al [20] investigated the taskoffloading problem by jointly considering the allocation ofradio resources and computational resources and proposedan iterative algorithm to solve the problem Beck et al [21]studied the virtual network embedding problems and pro-posed network virtualization in the context of MECnetworks

However the abovementioned methods still have somelimitations in performance and flexibility Furthermore thecomplexity of the algorithm is not suitable for large-scalenetwork scenarios -us an efficient task offloading mech-anism is designed formobile-edge cloud computing It is wellknown that the base stations in most wireless networks are

running under multichannel settings If a considerably largenumber of mobile device users simultaneously choose thesame wireless channel to perform task offloading interactivecommunication interference will seriously affect the trans-mission rate of the data which further leads to the increase ofcompletion delay of the task and the energy consumption ofmobile devices In this circumstance the task offloadingoperation violates the original intention of the mobile usersand these mobile users prefer local processing -e above-mentioned effect makes the offloading decision of mobiledevice users couple with the wireless resource allocation-us to achieve efficient taskoffloading two important issuesneed tobesolvedWhich tasksare suitable foroffloadingwhileothers are suitable for local execution How to choose theappropriate channel performing offloading from a globaloptimization

In order to deal with the abovementioned couplingproblems and match the mobility of the mobile device anenergy-aware task offloadingmechanism is designed to solvethe energy-efficient task offloading decision problem formobile-edge cloud computing in the multichannel wirelessenvironment -e mechanism adopts the auction theorywhich is a decentralized market mechanism of resource al-location in economy which has been widely used in variousfields nowadays such as the management of the spectrumresource in cognitive networks [22 23] and traffic offloadingin the cellular network [24 25] Inspired by the methods forresource allocation the auction is applied into the model ofwireless channel access opportunities for mobile devicesnamely task offloading decision In the auction system themobile device user acts as a buyer and the wireless channelsare treated as sellers -us the mobile device broadcasts theoffloading request and wireless channels will send their bidinformation with their available resources to the mobiledevice user -en the mobile device calculates the cost todetermine the winner channel to offload

-e contributions of this paper are listed as follows

(i) -e issueof the taskoffloadingdecision is formulatedas a constrained 0-1 nonlinear integer programmingproblem which minimizes the total energy con-sumption of the system subjected to the latency andchannel communication quality constrains

(ii) To reduce the complexity of solving the optimiza-tion problem an algorithm at first for classifyingmobile devices and priority determination is pro-posed -e mobile devices are classified into twogroups First category is suitable for local process-ing and the remaining mobile devices are classifiedas the second category which is determined by thecorresponding priority

(iii) In order to further determine the offloading de-cision for the mobile device an auction-based ap-proach is proposed to solve the energy-efficient taskoffloading decision problem with high efficiency toapproximate the optimal task offloading decisionsHence each mobile device can minimize its averageenergy consumption

2 Mobile Information Systems

e rest of this paper is organized as follows In Section 2the system model is described and the issue of task o-loading decision is formulated as a 0-1 nonlinear integerprogramming problem e details for solving the con-strained optimization problem and the proposed energy-aware task ooading mechanism are provided in Section 3In Section 4 the simulation results and the analysis of theresults are presented Finally concluding remarks are drawnin Section 5

2 System Model and Problem Formulation

In this section the mobile-edge cloud computing archi-tecture is presented and the 0-1 nonlinear integer pro-gramming problem is formulated for energy saving anddelay decreasing Table 1 lists all the important symbols usedin this paper

21 Mobile-Edge Cloud Computing Architecture As illus-trated in Figure 1 the mobile-edge cloud computing systemconsists of three key components mobile-edge cloud serverwireless access point and device users Mobile device usersrst connect to Internet through a base station which isequipped with a mobile-edge cloud server en the userscan further access this powerful mobile-edge cloud serverthrough the Internet After receiving the service request thecloud control provides the corresponding cloud services tothe mobile devices via the base station In this paper a set ofN 1 2 N devices are considered where each devicehas a task to be completed Moreover the wireless basestation runs M 1 2 M channel settings

e types of tasks include interactive gaming natural lan-guage processing virus scanning and video transcoding Eachtask is set in two terms asTiΔ(Oi Di) iisinN 1 2 N which canbe computed either locally on themobile device oron the cloud side via computation ooading It is assumedthat in the model each task is atomic and cannot be furtherdivided e characteristic information of all tasks can bedened through the total number of Oi CPU cycles re-quired to accomplish and the amount of Di data to beexchangede computation overhead is discussed in termsof both energy consumption and processing time for bothlocal and cloud computing approaches

For the local computing approach a mobile user i ex-ecutes the task Ti locally on the mobile device Taking intoaccount the dierent processing capabilities of dierentmobile devices as well as a dierent CPU computing ca-pacity per bit required for dierent tasks let Cli be thecomputation capability (ie CPU cycles per second) allo-cated by the mobile device i en the time duration of thelocal execution of the task Ti can be obtained as follows

tli OiCli (1)

e classic CPU energy consumption model of mobiledevices is applied [26 27] us the energy consumption Eliof this local execution can be calculated as

Eli (αi Cli( )χi + βi)t

li (2)

where the exponents χi αi and βi are the parameters whichare dependent on the CPU processing model All the energyconsumption parameters are set to conform to the actualmobile scenario in the subsequent simulation experiment

For the cloud computing approach once the com-munication link is established the task Ti on the mobile

Table 1 Important notations

Symbol DenitionN Mobile IoT device setM Wireless channel setTi e information of task on the device iOi e computing workload of the task iDi e data size for communication of the task iCli e computing capacity required for the task iCci e computation capacity allocated from the cloudPtri e data transmission power of the device iPidi e idle power of the device iGi e channel gainψij e ooading decisionwj e bandwidth of the channel jRij e transmission rate of the mobile device itli Local completion time of the task itci e cloud processing delay of the task iEli Local processing energy consumptionEtrij e data transmission power consumption of the task iEidi e idle power consumption of the mobile device i

EOffij

e total ooading energy consumption of the mobiledevice

Cloudcontroller

Data center

Internet

Health monitoringConnected vehicles

Smart mobile devices

MEC server

Figure 1 e architecture of mobile-edge cloud computing

Mobile Information Systems 3

device can be offloaded through the channel j jisinM tothe mobile-edge cloud server which will execute thecomputation task on behalf of the device user -e wholetask offloading process can be divided into three phrasesand involves corresponding time delay and energyconsumption

To begin with the mobile device transmits the dataload of the task i to the closest base station through thechannel j which involves the data transmission delay ttrijand energy consumption of the data transmission Etr

ij ofthe task -e energy consumption of the data transmissionis not only related to the data size but also to the uplinkdata rate -us the binary variable ψij isin 0 1 is denotedas the task offloading decision of the mobile user i which isgiven by

ψij 1 offload the task of device i via channel j

0 process task of device i locally1113896

(3)

where ψij 1 means that the device i chooses to offload thetask and transmits the computation data to the cloudthrough the channel j jisinM while ψij 0 denotes that thedevice executes the task locally According to the offloadingdecisions an appropriate number of VMs are deployed inthe data servers for cloud execution Moreover given theoffloading decisions Ψ (ψ1jψ2j ψNj) of all deviceusers the uplink data rate Rij of a device user i who choosesto offload the task to the cloud via the wireless channel j canbe calculated as [19]

Rij(Ψ) wjlog2 1 +Ptr

i Gi

σ2 +sumrisinN i ψij1Ptrr Gr

⎛⎝ ⎞⎠ (4)

where wj is the j channel bandwidth and Ptri represents

the data transmission power of mobile devices which isdetermined by the wireless base station according tosome power control algorithms such as [28] and [29]Furthermore Gi ℓminusai denotes the channel gain betweenthe device user i and the base station where ℓi indicatesthe distance between the mobile device i and the wirelessbase station Moreover σ2 denotes the background noisepower and the parameter a denotes the path loss factor-erefore the data transmission delay ttrij can be calcu-lated as

ttrij

Di

Rij

(5)

Let Ptri be the data transmission power consumption

-e energy consumption of the data transmission can bedenoted as

Etrij

Di

Rij

Ptri (6)

In the second phase the base station transmits the dataload to the mobile-edge cloud server through a high-speedwire network [27 28] Because of the high-speed linkthe time delay of this phase can be ignored In the final phase

the mobile-edge cloud server processes the task and returns theresults back to the device users Because the size of results isoften considerably smaller than that of the input data loadthe time delay from the mobile-edge cloud server to thedevice is not considered as in some of the previous studies-erefore the delay of this phase is mainly composed of thecloud processing delay tc

i Let Cci be the computation ca-

pacity allocated from the cloud-e cloud processing time tci

spent on the cloud side can be formulated by

tci

Oi

Cci

(7)

When the task is executed on the cloud the mobiledevice needs to wait for the return of the response result-us at the period of time tc

i spent on the task processing onthe cloud side the idle power consumption of the mobiledevice can be calculated as follows

Eidi

OiPidi

Cci

(8)

-us the total offloading energy consumption EOffij is

expressed as EOffij Etr

ij + Eidi which is defined as the sum of

the energy spent to transmit data to the cloud Etrij plus the

idle power consumption Eidi as follows

EOffij

DiPtri

Rij

+OiP

idi

Cci

(9)

And the total time delay of offloading is denoted asTOff

ij tci + ttrij which is defined as the transmission delay

plus the cloud processing delayIn order to realize energy-efficient task offloading it is

necessary to properly deal with the reasonable allocation ofcommunication resources and computing resources whichare mutually coupled in the case of energy efficiency becauseof the competition for the resource However in this paperit is assumed that the processing capacity of cloud services isfar greater than the processing capacity of each mobiledevice and the computing and storage resources are suffi-cient to satisfy the requirements of all mobile devices Be-sides our research scope of this paper is mainly focused onthe mobile devices for the purpose of saving energy byoffloading the task onto the cloud side -erefore the serveroverhead and energy consumption of the cloud server arenot considered which does not affect the completeness ofthe paper

Among the multiple mobile device users for the mobile-edge cloud computing environment the mobile device se-lects the nearest wireless access point in order to get bettercommunication and interaction Similar to the previousresearch on mobile-edge cloud computing from the be-ginning of the offloading decision until the end of offloadingoperation it is reasonable to assume that all mobile devicesmove very slowly in a quasi-static scenario

Tominimize the total energy consumption of the systemthe optimization problem is mathematically modeled asfollows


minψij1113864 1113865

F sumN

i11minusψij1113872 1113873E

li + ψijE

Offij1113874 1113875 (10)

st sumM

j1ψijE

Offij leE

liforalliisinN (11)

sumM

j1ψijT

Offij leT

liforalliisinN (12)

sumN

i1ψijP

tri Gi leCforalljisinM (13)

sumM

j1ψij le 1foralliisinN (14)

ψij isin 0 1 foralliisinNforalljisinM (15)

Constraint (11) ensures that the energy consumption oftask offloading is not greater than the local processing energyconsumption of the mobile device Constraint (12) ensuresthat the total time consumption of mobile devices in theprocess of task offloading is not greater than the localprocessing energy consumption of the task Constraint (13)is to guarantee the communication quality of the wirelesschannel -e setting of the threshold C can avoid mobiledevices to access the same channel at the same time becausethe burst data traffic of mobile devices will seriously cause theattenuation of channel quality Constraint (14) states thatthemobile device can only select access to a wireless channelbut the wireless channel can be accepted by a plurality ofmobile devices Constraint (15) states that the cloud off-loading decision of the task is a binary variable

Consider that the task offloading decisions Ψ among thedevice users are coupled If too many device users simulta-neously choose to offload the task to the cloud via the samewireless channel they may cause severe interference whichwill lead to a low data rate -e two factors related to theenergy consumption of data transmission of themobile deviceare the inherent transmission power and data transmissiontime-e transmission energy consumption of mobile devicesis proportional to the transmission time -us when the datarate of the mobile device user is low it would consume highenergy and incur long transmission time as well In this casemore and more device users will avoid offloading and aremore willing to choose execution locally However this is notour original intention allowing beneficial cloud computingusers to offloading as much as possible-us theC thresholdis set which can be flexible assignment

3 Energy-Aware Task Offloading Mechanism

To solve the optimization problem (10) an energy-awaretask offloading mechanism is designed in the system ofmobile-edge cloud computing -e proposed mechanismmainly includes two aspects

(1) At the beginning an algorithm for mobile deviceuser classification and priority determination isdesigned -e mobile device users can be classifiedinto two types participation in the auction and not

to participate according to the energy cost features ofthe task computing process Namely the mobiledevice users who do not participate in the auctionchoose to process the task locally-en the prioritiesof the first class of users are derived which representthe intensity of user demand for task offloading

(2) According to the order of priority the device usersget resource allocation in turn A reverse auction-based offloading algorithm is proposed to achieve theoffloading decision and associate the suitable com-munication resource with each mobile device whoparticipates in the auction

31 Mobile Device User Classification and PriorityDetermination Based on the characteristics of the task andthe mobile device such as the data size of the task workloaddensity computing capacity and energy consumption themobile device users are divided into two types

-e first type of users is a group that should computetheir task locally -e set of users of this type is denotedas Gl When the mobile device occupies a channel alonethe data transmission rate of this mobile device can beexpressed by

R0ij wjlog2 1 +

Ptri Gi

σ21113888 1113889 (16)

-e condition used to determine the devices belongingto this type is given as follows

Theorem 1 if Eli ltE

Offij iisinN jisinM then the device i be-

longs to Gl where

EOffij

DiPtri

R0ij

+OiP

idi

Cci

(17)

Besides the aforementioned type the rest of the mobiledevice users fall into second typeGo-emobile device usersbelonging to Go can either decide to implement their tasklocally or to offload the task onto the mobile-edge cloudserver-e decision of them depends on the communicationquality of the channel For this type of mobile users differentpriorities are set for them in the offloading process which isdefined as

ηi GiP

tri

Eli

1113969 (18)

-e complete mobile device user classification andpriority determination are illustrated in Algorithm 1

32 Reverse Auction-Based Offloading Algorithm In thissection a reverse auction-based offloading scheme is pro-posed for the group Go of mobile devices based on theabovementioned analysis of mobile device user classificationand priority determination Our aim is to maximize theenergy efficiency of task offloading subjected to the mobiledevicersquos minimum energy consumption requirement and


the limited communication resource of channels during thetask ooading process

As illustrated in Figure 2 the mobile device i iisinGo actsas the buyer who achieves higher system energy eciency inexchange of transmission power resources provided by thechannel Prior to participation in the auction namely de-ciding whether to ooad the task onto the mobile-edgecloud server mobile users rst calculate their costPi whichmeans the reserved prices the mobile device can accept Inthis case the reserve price corresponds to the aforemen-tioned local computing energy consumption of the task inthe system of mobile-edge cloud computing us the re-serve price Pi is expressed as Pi Eli

On the other hand the wireless channels are sellers Eachchannel can participate in the auction by submitting to themobile device the bidding information (bj sj) where bj isthe price at which the jth channel agrees to share theiravailable resources to the mobile device Each seller calcu-lates their sj and bj respectively by sj Ptr

i Gij andbj EOff

ij e total number of available resources of eachseller corresponds to the aforementioned interferencethreshold of each channel en the mobile device willcalculate the energy cost and compare the biding pricesprovided by the seller to decide whether it could achieveenergy saving and choose the target channel or give up taskooading decision e target of the wireless channel is thewinner of the reverse auction process e mobile devicechooses the target channel for task ooading

In the previous auction researches they allocate theresources through multiround bidding procedures to de-termine the nal winner However this multiround auctionmethod is not suitable for our scenario because mobiledevice users have to wait for the consequences after multiplerounds of auction which inevitably generate an intolerable

extra delay In the process of task ooading mobile devicesare sensitive to delay erefore the single-round auction isimplemented in this paper in order to improve the energyeciency and reduce the delay for the ooading usersMoreover it is assumed that the time delay of the auctionprocess is so small that it can be ignored e auction isconducted periodically which means that after a smaller

InitializationMobile IoT device set N 1 2 N Wireless channel set M 1 2 M e task on mobile IoT device TiΔ(Oi Di)Transmission power of mobile IoT device Ptr

i i isinNIdle power of mobile IoT device Pid

i i isinNCategorized device sets Gl Go emptyPriority set η empty1 for mobile device i 1 to N do2 for channel j 1 to M do3 calculate the exclusive channel data transfer rate R0

ij ofeach mobile device as in (11) and the energy consumption EOff

ij as in (12)4 if Eli leEOff

ij then5 irArrGl6 else7 irArrGo8 ηi GiPtr

i Eli

radic

9 end if10 end forOutpute categorized device set Gl Goe priority set for the devices η ηi i isin Go

ALGORITHM 1 e Algorithm for classifying the mobile device and priority determination

Seller 1 Seller 2 Seller 3

Buyer Mobile device

Wireless channel

Sending bid informationRequesting for offloading

Bids collection

AuctionAllocation

Pricing

Response

helliphellip

Figure 2 e reverse auction system e mobile device sends therequestoftransmissionwiththeenergycostcollectsthebidinformationsent by wireless channels and then chooses the winner channel


auction interval a new round of auction is started and therelevant information is collected again which is adapted tothe dynamic mobile cloud computing environment In orderto simplify the model it is assumed that the auction intervalis very short and is ignored -e complete reverse auction-based offloading algorithm is illustrated in Algorithm 2

321 Allocation In the allocation steps the mobile devicedecides which channel will be the auction winner In order toavoid the extra delay caused by the multiround auction thesingle-round auction is implemented in this paper -usjointly considering the resources and the price that thebidders can provide the mobile device decides who will winthe auction and bj is the transaction price -erefore giventhe abovementioned definitions and notation the optimi-zation problem can be converted into the reverse auctionproblem Here ψij represents the consequence of auctionψij 0 denotes that there is no winner channel On thecontrary ψij 1 expresses that the jth channel wins theauction Our goal is to maximize the utility of the mobiledevice user which can be formulated as

maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN

i11113874 1minusψij1113872 1113873Pi + ψijbj1113875 (19)

In order to determine the winner and the allocationrelationship the bid densities of the participants are cal-culated and sorted firstly In the list of wireless channels thewireless channels were ranked in ascending order of theirbid densities For mobile users the lowest call density is the

best communication quality-e bid density of sellers can becalculated by

bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij

Cj minussumNr1rne iψrjP

tri Gi

1113969 (20)

where Cj minussumNr1rne iψrjP

tri Gi gt 0 which is an indispensable

condition for the wireless channel to ensure their quality ofservice If the value is less than or equal to zero the channelwill give up participating in the auction

322 Pricing Model -e final transaction price paid by themobile device is bj which is the bid price submitted by thewinner wireless channel-e utility of the mobile device usercan be formulated as

F sumN

i1Pi minus sum

M

j1sumN

i11minusψij1113872 1113873Pi + ψijbj1113874 1113875 (21)

If the mobile user does not participate in the auction itsutility value is equal to 0 In other words if ψij 0 obviouslythen F 0 through the calculation of formula (21) More-over the utility of the wireless channel can be formulated as

Θ sumM

j1sumN

i1ψijbj (22)

If the wireless channel does not win the auction thenψij 0 obviously the utility of the wireless channel is equalto zero

Input Gl Go ηOutput Offloading decision Ψ (ψ1jψ2j ψNj)1 Set the temporary set Go

prime Go2 while Go

prime neempty do3 Select the device i where i argmax ηi1113864 1113865i i isin Go4 for channel j 1 to M do5 Update the data transmission rate Rij and update EOff

ij as in (4) and (9)6 if Cj gt 0 then7 Calculate the bid density bdj of each channel j based on the 2-tuple (bj sj)8 Set the bid density bd bdj1113966 11139679 while bdneempty do10 Select the channel j where j argmin bdj1113966 1113967

j

11 if EOffij leEl

i ampamp Cj minussumNr1rne iψrjPtr

i Gi gt 0 then12 Let ψij 1

13 CjlArrCj minussumNr1rne iψijPtr

i Gi14 else15 Let ψij 016 end if17 bd bdj18 end while19 else20 Let ψij 021 end if22 end for23 Go

prime Goprime i

24 end while

ALGORITHM 2 Reverse Auction-Based Offloading Algorithm for Offloading Decisions


323 Properties In this section the properties of theproposed reverse auctionmodel are analyzed-e individualrationality and the truthfulness properties need to be proved

(1) Individual rationality when the utility of each par-ticipating bidder in the pricing stage is greater thanzero the proposed mechanism is individual rationalfor each winning bidder Namely

FBM bj1113864 1113865minus FBM

minusF bj1113874 1113875ge 0 (23)

where BM b1 b2 bj bM1113966 1113967 and FBM bj1113864 1113865

denotesthe utility of the mobile device under the optimal allocationsolution without the presence of the jth channel

(2) Truthfulness for each bidder the truthfulness meansthat the bid price of each bidder is equal to its privatevalue If the bidding of channels is untrue the utilitywill be unlikely the biggest In order to get the max-imum the allocation should be formulated as follows


minusF bj1113874 1113875ge 0 (24)

Ω FBM bj1113864 1113865minus FBM

minusF bj1113874 1113875minus FBM bi minus FBM

minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865

+ F bj1113874 1113875minus FBM bi + F bi

1113874 1113875

(25)

Based on the proposed reverse auction mechanism in thispaper because the bid price of the channel is not greater thanthe reverse price of the mobile device user Ωle 0 Obviouslywhen j i the value of Ω is equal to zero -erefore eachbidder must be truthful to obtain the maximum utility

4 Simulation and Analysis

In this section the performance of the proposedmechanism isevaluated through numerical simulations designed by usingtheMATLAB-e compared algorithms are the competition-based algorithm [30] and the user-satisfaction-based off-loading algorithm [31]-eir features are described as follows

(1) Competition-based algorithm the system is modeledas a competitive game subjected to the job executiondeadlines and user-specific channel bit rates Eachuser tries to minimize its own energy consumptionwhen it competes for the shared communicationchannel -e GaussndashSeidel-like method is executedfor achieving the Nash equilibrium to derive themobile device userrsquos offloading decisions

(2) User-satisfaction algorithm a utility function is in-troduced to choose the best communication resourcesin terms of user-satisfaction parameters such as thethroughput used energy and time spent to execute theapplication Based on this the offloading strategy isobtained by the applicationsrsquo computation percentage

Without loss of generality four performance metrics ofthe proposed algorithm and the two classical algorithms are

compared on the same simulation scenarios fairly -e fourmetrics are the average energy consumption delay energyefficiency factor and throughput of the mobile device foroffloading

41 Simulation Setup -e simulations are deployed basedon real-world settings All the parameters including theenergy consumption rates and computing capacity aremeasured from real mobile devices -ese real-worlddatasets which have been widely used are measured atvarious clock speeds and in the cellular network scenarios byusing a monsoon power monitor

At first a base station is considered that covers a hex-agonal cellular network with radius 2 km and assume thatthe wireless access point is located at the center -e basestation has M 4 channels and the channels belonging tothis base station are orthogonal -e bandwidth capacity ofthe channels can be different values but in order to simplifythe simulation four channels of the same bandwidth of thedevice are set to w 1MHz which does not affect the effectof the experiment Besides the power of the backgroundnoise is set to σ2 minus100dBm and the path loss factor is set toa 2 according to the physical interference model In thesystem of mobile-edge cloud computing mobile devices arerandomly distributed in the coverage area of the hexagonalcellular network accessing to this wireless point at any timeif needs And there is a mobile-edge server deployed near thebase station who assigns 5GHz computation capability foreach mobile device sufficient to satisfy the requirements ofall mobile devices

Conforming to the diversity of the mobile device in thereal world four types of smartphones are considerednamely Galaxy Note Galaxy Note 2 Nexus S and HP iPAQPDA Different mobile devices have different CPU com-puting capacities -e HP iPAQ PDA with a 400MHz IntelXScale processor [31] has the following parameters the localprocessing power Pl

i 09W the standby power Pidi 03W

and the transmission power Ptri 13W In addition the

parameters of the other three mobile devices include CPUprocessing parameters such as χi αi and βi -ese pa-rameters are adopted as in [30] In the simulation the type ofthe mobile device in the mobile-edge cloud computingscenario is randomly selected among the abovementionedthree types and eachmobile device has only one task waitingto be executed -e tasks on mobile devices are set to tentypes face recognition virus scanning online gaming andso on-ese ten types of tasks are randomly assigned to eachmobile user Different types of mobile devices have differentprocessing speeds for different task types whose corre-sponding parameters are given in Table 2 includingworkload density data size and the allocated computingcapacity

It is clear that in these tasks the workload densities offace recognition and virus scanning are larger than those ofother types of tasks and the data size of the two tasksis relatively small which are computation-intensive tasksOn the contrary the workload density of video coding is farless than that of the other eight tasks but the data size is


particularly larger than that of others which belong tocommunication-intensive tasks It is obvious that the pa-rameters given in the table include various types of tasks thatsatisfy the generality and credibility of the simulation

In order to accurately evaluate the performance of thealgorithm without any loss of generality a series of simu-lations are carried out gradually increasing the number ofmobile devices from 50 to 1000 Since the mobile devices arerandomly deployed within the coverage of mobile networksand the type of the mobile device and the task request ofmobile users have stochastic features

42 Evaluation Results Firstly the energy consumption ofthe proposed algorithm is evaluated Figure 3 shows theaverage energy consumption of the mobile device when thenumber of mobile devices increases from 50 to 1000 withfour dierent methods e average energy consumed byone mobile device is approximately 212060 J with the localcomputing approach Comparing with the local computingapproach both the proposed approach and the other twoalgorithms achieve the purpose of energy saving throughtask ooading

At the beginning with 50 mobile devices the threemethods exhibit an energy consumption of 64219 J 64276 Jand 69430 J respectively With the gradually increasednumber of mobile devices the average energy consumptionof the mobile device increases to 110077 J 125876 J and133540 J respectively is is because too many mobiledevices choose to access the same wireless channel to im-plement the task ooading simultaneously which wouldlead to the augment of mutual interference According to (4)it is obvious that the severe interference to each other willcause the reduction of the communication quality and therates for computation ooading erefore with 1000mobile devices more andmore users tend to choose the localcomputing method and the average energy consumption of

mobile devices increases e proposed mechanism can saveat least 5642 of the energy consumption

e superiority of the proposed approach is graduallyobviousis is due to the fact that the reverse auction-basedooading mechanism performs task ooading decision ina global long-term perspective reasonably allocating com-munication resources for mobile device users to meet thequality of service requirements It exhibits a relatively lowerenergy consumption when the number of mobile deviceusers is small However with the explosive increase in the

Table 2 Parameters of the system

Smartphone χi αi βi Ptri Pid

i

Galaxy Note 30 033 010 2605 964Galaxy Note 2 27 025 040 2796 1170Nexus S 30 034 035 1217 74Galaxy Nexus 30 040 030 964 2237Task Oi Di Cli mdash mdashFace recognition 60 31680 12 mdash mdash400-frame game 2048 2640 10 mdash mdashChess select 400 1580 06 mdash mdashChess move 400 2640 10 mdash mdashVirus scanning 300 32946 15 mdash mdash4-queen puzzle 200 878 04 mdash mdash5-queen puzzle 200 263 045 mdash mdash6-queen puzzle 200 1760 072 mdash mdash7-queen puzzle 200 8250 104 mdash mdashVideo transcoding 10240 200 056 mdash mdash

50 200 500 800 10000

5

10

15

20

25

30

35

Number of mobile devices

Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)

The proposed algorithmCompetition-based algorithm [30]User-satisfaction algorithm [31]All tasks are executed locally

Variance

Figure 3 Average energy consumption of mobile devices


number of mobile device users the performance degradesdue to the trac growth Obviously the proposed methodcan nd a better energy-saving solution than other twoapproaches

e average task execution delay of mobile devices withthe proposed method and the other two schemes is com-pared As shown in Figure 4 the average time delay ofmobile devices for performing a task is approximately275157 s with the local computing approach With 50mobile devices the time delays of these three methods are146868 s 146605 s and 155306 s respectively Comparedwith the local computing approach at least 4569 of thetime delay can be saved When 1000 mobile devices aredeployed the time delays obtained by four methods are196782 s 200610 s 208447 s and 275751 s respectivelye proposed mechanism can save about 3543 of the timecompared to the local computing approach which is slightlyhigher than the performance of other two algorithms

ird the throughput of mobile devices is compared inthe case of our proposed method and other two methods inaddition to the local computing method because the localprocessing does not need to upload data and the throughputis zero Figure 5 shows that at the beginning with 50 mobiledevice users the other three methods exhibit an averagethroughput of 51446 bps 58837 bps and 61957 bps re-spectively Although the throughput of mobile devices in thecase of our proposed algorithm is lower at the beginningwhen the number of mobile devices is between 50 and 200the trend of throughput drops more slowly than the othertwo methods With the continued growth of mobile devicesthe throughput of mobile devices in the case of the proposedmethod is higher than that in the other methods At the endwith 1000 mobile device users the methods exhibit an

average throughput of 9323 bps 7964 bps and 8195 bpsrespectively As the number of mobile devices increases log-arithmically the correspondingly mutual interference amongthe device will grow Furthermore the uplink data trans-mission ratewill decreasewhich leads theenergyconsumptionof cloud ooading greater than that of local computingusmore and more mobile device users will adopt local com-puting substituting for ooading operation Comparedwiththe competition-based algorithm and the user-satisfactionalgorithm the throughput is higher and the rate of decline isrelatively slow when using the proposed method

Finally the energy eciency factor for ooading isevaluated with the proposed method competition-basedalgorithm and user-satisfaction algorithm over 1000 sim-ulation runse proposed mechanism is designed to reducethe energy consumption and the response time delay ofmobile devices us a function is proposed representingthe QoS degree perceived by the user e function ismodeled as a sigmoid curve which is widely used to measureuser satisfaction and service quality in previous studies [32]User satisfaction increases as energy consumption and la-tency decrease so we use sigmoid functions to representthe relationship between them f1 1minus (11 + eminus(EaverminusEl))f2 1minus (11 + eminus(Taverminustl))

e function U ω1f1 + ω2f2 is introduced to analyzethe energy eciency factor where ω1 + ω2 1 Moreoverω1 and ω2 represent the weight coecients of energy con-sumption and delay respectively And Eaver and Taver re-spectively denote the average energy consumption andaverage delay As shown in Figure 6 with an increasednumber of mobile devices the user satisfaction for taskooading gradually reduced And at the last with 1000mobile devices the values of the two methods of comparisondrop sharply On the contrary compared with the other twocurves the curve corresponding to the proposed method isrelatively stable erefore when there are a large number of

102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)

The proposed algorithmCompetition-based algorithm [30]User-satisfaction algorithm [31]

Figure 5 Average throughput of mobile devices

0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance

50 200 500 800 1000Number of mobile devices


Figure 4 Average delay of mobile devices


mobile devices in the mobile-edge cloud system the per-formance of the proposed method is better to meet therequest of mobile device users for cloud ooading

5 Conclusion

In this paper an energy-aware task ooading mechanism isdesigned to perform ooading decisions with optimizationon minimizing the energy consumption of mobile devicesConsidering the interference threshold in each channel thetask local execution delay and the local energy consump-tion the task ooading decision problem is formulated asa 0-1 nonlinear integer programming optimization In orderto solve this problem the algorithm is proposed for clas-sifying the mobile device and priority determination Fur-thermore the reverse auction theory has been implementedwith the proposed algorithm to decide the ooading targetchannel e individual rationality and truthfulness of thereversed auction model are also discussed in the paper eperformances of the proposed mechanism comparing withthe other two methods are evaluated with performancemetrics of energy consumption time delay throughout andthe energy eciency factor e simulation results validatethat the proposed algorithm can achieve better performances

Conflicts of Interest

e authors declare that they have no consecticts of interest

Acknowledgments

is work was partially supported by the National NaturalScience Foundation of China (Grant nos 6137911161402538 61403424 61502055 61672537 and 61672539)

References

[1] R Janessa and R Meulen Gartner Says the Internet of ingsInstalled Base Will Grow to 26 Billion Units by 2020 GartnerInc Stamford CT USA 2013

[2] CISCO e Internet of ings How the Next Evolution of theInternet is Changing Everything CISCO White Paper 2011

[3] E Ahmed A Gani M K Khan R Buyyac and S U KhanldquoSeamless application execution in mobile cloud computingmotivation taxonomy and open challengesrdquo Journal ofNetwork and Computer Applications vol 52 pp 154ndash1722015

[4] H T Dinh C Lee D Niyato and P Wang ldquoA survey ofmobile cloud computing architecture applications and ap-proachesrdquo Wireless Communications and Mobile Computingvol 13 no 18 pp 1587ndash1611 2013

[5] S Barbarossa S Sardellitti and P D Lorenzo ldquoCommuni-cating while computing distributed mobile cloud computingover 5G heterogeneous networksrdquo IEEE Signal ProcessingMagazine vol 31 no 6 pp 45ndash55 2014

[6] S Abolfazli Z Sanaei E Ahmed et al ldquoCloud-based aug-mentation for mobile devices motivation taxonomies andopen challengesrdquo IEEE Communications Surveys and Tutorialsvol 16 no 1 pp 337ndash368 2014

[7] X Chen ldquoDecentralized computation ooading game formobile cloud computingrdquo IEEE Transactions on Parallel andDistributed Systems vol 26 no 4 pp 974ndash983 2014

[8] S Barbarossa S Sardellitti and P D Lorenzo ldquoJoint al-location of computation and communication resources inmultiuser mobile cloud computingrdquo in Proceedings of theIEEE 14th Workshop on Signal Processing Advances inWireless Communications pp 26ndash30 Darmstadt GermanyJune 2013

[9] D Huang P Wang and D Niyato ldquoA dynamic ooadingalgorithm for mobile computingrdquo IEEE Transaction onWireless Communications vol 11 no 6 pp 1991ndash1995 2012

[10] L Yang J Cao H Cheng and J Yusheng ldquoMulti-usercomputation partitioning for latency sensitive mobile cloudapplicationsrdquo IEEE Transactions on Computers vol 64 no 8pp 2253ndash2266 2015

[11] H Viswanathan E K Lee I Rodero and D PompilildquoUncertainty-aware autonomic resource provisioning formobile cloud computingrdquo IEEE Transactions on Parallel andDistributed Systems vol 26 no 8 pp 2363ndash2372 2015

[12] O Munoz-Medina A Pascual-Iserte and J Vidal ldquoOpti-mization of radio and computational resources for energyeciency in latency-constrained application ooadingrdquo IEEETransactions on Vehicular Technology vol 64 no 10pp 4738ndash4755 2015

[13] M Satyanarayanan P Bahl R Caceres and N Davies ldquoecase for VM-based cloudlets in mobile computingrdquo IEEEPervasive Computing vol 8 no 4 pp 14ndash23 2009

[14] Y Zhang D Niyato and P Wang ldquoOoading in mobilecloudlet systems with intermittent connectivityrdquo IEEETransactions on Mobile Computing vol 14 no 12 pp 2516ndash2529 2015

[15] W Li Y Zhao S Lu and D Chen ldquoMechanisms andchallenges on mobility-augmented service provisioning formobile cloud computingrdquo IEEE Communications Magazinevol 53 no 3 pp 89ndash97 2015

[16] L Lei Z Zhong K Zheng J Chen and H Meng ldquoChallengeson wireless heterogeneous networks for mobile cloud com-putingrdquo IEEE Wireless Communications vol 20 no 3pp 34ndash44 2013

102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


Figure 6 Energy eciency factor of mobile devices


[17] Y Mao C You J Zhang K Huang and K B LetaiefldquoA survey on mobile edge computing the communicationperspectiverdquo IEEE Communications Surveys amp Tutorialsvol 99 2017

[18] S Wang R Urgaonkar M Zafer and T He ldquoDynamicservice migration inmobile edge-cloudsrdquo in Proceedings of theIFIP Networking Conference pp 1ndash9 Toulouse France March2015

[19] X Chen L Jiao W Li and X Fu ldquoEfficient multi-usercomputation offloading for mobile-edge cloud computingrdquoIEEEACM Transactions on Networking vol 24 no 4pp 974ndash983 2015

[20] S Sardellitti G Scutari and S Barbarossa ldquoJoint opti-mization of radio and computational resources for multicellmobile-edge computingrdquo IEEE Transactions on Signal andInformation Processing Over Networks vol 1 no 2 pp 89ndash103 2015

[21] M T Beck and M Maier ldquoMobile Edge Computing Chal-lenges for Future Virtual Network Embedding Algorithmsrdquogte Eighth International Conference on Advanced EngineeringComputing and Applications in Sciences pp 65ndash70 RomeItaly 2014

[22] Y Zhang C Lee D Niyato and P Wang ldquoAuction ap-proaches for resource allocation in wireless systems a surveyrdquoIEEE Communications Surveys and Tutorials vol 15 no 3pp 1020ndash1041 2013

[23] B Kollimarla Spectrum Sharing in Cognitive Radio College ofOklahoma State University Oklahoma City OK USA 2009

[24] G Iosifidis L Gao J Huang and L Tassiulas ldquoA double-auction mechanism for mobile data-offloading marketsrdquoIEEEACM Transactions on Networking vol 23 no 5pp 1634ndash1647 2015

[25] S Paris F Martignon I Filippini and L Chen ldquoAn efficientauction-based mechanism for mobile data offloadingrdquo IEEETransactions on Mobile Computing vol 14 no 8 pp 1573ndash1586 2015

[26] J Kwak Y Kim J Lee and S Chong ldquoDREAM dynamicresource and task allocation for energy minimization inmobile cloud systemsrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 12 pp 2510ndash2523 2015

[27] K Son and B Krishnamachari ldquoSpeedBalance speed-scaling-aware optimal load balancing for green cellular networksrdquo inProceedings of the IEEE INFOCOM 2012 pp 2816ndash2820Orlando FL USA March 2012

[28] M Xiao N B Shroff and E K P Chong ldquoA utility-basedpower-control scheme in wireless cellular systemsrdquoIEEEACM Transactions on Networking vol 11 no 2pp 210ndash221 2003

[29] M Chiang P Hande T Lan and C W Tan ldquoPower controlin wireless cellular networksrdquo Foundations and Trends inNetworking vol 2 no 4 pp 381ndash533 2008

[30] E Meskar T Todd D Zhao and G KarakLondon UKostasldquoEnergy efficient offloading for competing users on a sharedcommunication channelrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC) pp 3192ndash3197 London UK June 2015

[31] D Mazza D Tarchi and G E Corazza ldquoA user-satisfactionbased offloading technique for smart city applicationsrdquo inProceedings of the 2014 IEEE Global CommunicationsConference pp 2783ndash2788 Austin TX USA December2014

[32] D H V Seggern CRC Standard Curves and Surfaces withMathematica CRC Press Boca Raton FL USA 2015


Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems


Volume 2018


ReconfigurableComputing



Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018


Civil EngineeringAdvances in


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia


Biomedical Imaging



Engineering Mathematics


RoboticsJournal of



Computational Intelligence and Neuroscience


Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018


Human-ComputerInteraction

Advances in


Scientic Programming

Submit your manuscripts atwwwhindawicom

users and designed an offline heuristic offloading algorithmViswanathan et al [11] proposed a resource provisioningframework for organizing the heterogeneous devices in thevicinity A joint optimization framework of the wirelessresource and computing resource is proposed in [12] for theenergy-constrained mobile users in the femtocell

In order to solve the problem of transmission delay andenergy consumption Satyanarayanan et al [13] proposed anarchitecture replacing the remote cloud with nearby cloud-lets Zhang et al [14] proposed a Markov decision processoffloading algorithm for mobile users in an intermittentlyconnected cloudlet system However the cloudlet-basedmobile cloud computing has some drawbacks Due to thelimited wireless network coverage the cloudlet cannotguaranteeubiquitous service everywhere forusersMoreoverthe computation resource of the cloudlet is insufficient tosatisfy theQoS requirements of a large number of users in thefuture

-ese existing task offloading strategies in mobile cloudcomputing are not sufficient to greatly improve the energyefficiency of the system-e long transmission delay betweenthe mobile device and traditional cloud servers is a criticalissue which is the inherent limitation of mobile cloudcomputing [15 16] -e long propagation distance from themobile device to the remote cloud data center will result inunacceptable long latency for mobile applications -e ad-ditional communication transmission delay will decrease thecomputation offloading efficiency and the QoS of users

-erefore a mobile-edge cloud (MEC) computing ar-chitecture is adopted With the development of wirelesscommunication technologies such as Wi-Fi 4G and 5G theMEC is envisioned as a promising and challenging approachto address the abovementioned challenges [17] In the MECframework mobile devices are able to offload their tasks tothe MEC clouds through the radio access points nearbyrather than the public clouds such as Amazon EC2 andWindows Azure -us this MEC paradigm can providelower latency and high communication rate and computingagility in the process of computation offloading

-ere are a few studies on the efficient computationoffloading mechanism of the MEC For instance Wang et al[18] used the Markov decision process to formulate a se-quential offloading decision-making problem for dynamicservice migration Considering the finite number of wirelessaccess channels and the interference Chen et al [19] pre-sented a distributed offloading decision method based ongame theory Sardellitti et al [20] investigated the taskoffloading problem by jointly considering the allocation ofradio resources and computational resources and proposedan iterative algorithm to solve the problem Beck et al [21]studied the virtual network embedding problems and pro-posed network virtualization in the context of MECnetworks

However the abovementioned methods still have somelimitations in performance and flexibility Furthermore thecomplexity of the algorithm is not suitable for large-scalenetwork scenarios -us an efficient task offloading mech-anism is designed formobile-edge cloud computing It is wellknown that the base stations in most wireless networks are

running under multichannel settings If a considerably largenumber of mobile device users simultaneously choose thesame wireless channel to perform task offloading interactivecommunication interference will seriously affect the trans-mission rate of the data which further leads to the increase ofcompletion delay of the task and the energy consumption ofmobile devices In this circumstance the task offloadingoperation violates the original intention of the mobile usersand these mobile users prefer local processing -e above-mentioned effect makes the offloading decision of mobiledevice users couple with the wireless resource allocation-us to achieve efficient taskoffloading two important issuesneed tobesolvedWhich tasksare suitable foroffloadingwhileothers are suitable for local execution How to choose theappropriate channel performing offloading from a globaloptimization

In order to deal with the abovementioned couplingproblems and match the mobility of the mobile device anenergy-aware task offloadingmechanism is designed to solvethe energy-efficient task offloading decision problem formobile-edge cloud computing in the multichannel wirelessenvironment -e mechanism adopts the auction theorywhich is a decentralized market mechanism of resource al-location in economy which has been widely used in variousfields nowadays such as the management of the spectrumresource in cognitive networks [22 23] and traffic offloadingin the cellular network [24 25] Inspired by the methods forresource allocation the auction is applied into the model ofwireless channel access opportunities for mobile devicesnamely task offloading decision In the auction system themobile device user acts as a buyer and the wireless channelsare treated as sellers -us the mobile device broadcasts theoffloading request and wireless channels will send their bidinformation with their available resources to the mobiledevice user -en the mobile device calculates the cost todetermine the winner channel to offload

-e contributions of this paper are listed as follows

(i) -e issueof the taskoffloadingdecision is formulatedas a constrained 0-1 nonlinear integer programmingproblem which minimizes the total energy con-sumption of the system subjected to the latency andchannel communication quality constrains

(ii) To reduce the complexity of solving the optimiza-tion problem an algorithm at first for classifyingmobile devices and priority determination is pro-posed -e mobile devices are classified into twogroups First category is suitable for local process-ing and the remaining mobile devices are classifiedas the second category which is determined by thecorresponding priority

(iii) In order to further determine the offloading de-cision for the mobile device an auction-based ap-proach is proposed to solve the energy-efficient taskoffloading decision problem with high efficiency toapproximate the optimal task offloading decisionsHence each mobile device can minimize its averageenergy consumption








tli OiCli (1)



li (2)





EOffij


Cloudcontroller

Data center

Internet



MEC server








(3)



i Gi


⎛⎝ ⎞⎠ (4)



ttrij

Di

Rij

(5)



Etrij

Di

Rij

Ptri (6)






tci

Oi

Cci

(7)



Eidi

OiPidi

Cci

(8)






EOffij

DiPtri

Rij

+OiP

idi

Cci

(9)








minψij1113864 1113865

F sumN

i11minusψij1113872 1113873E

li + ψijE

Offij1113874 1113875 (10)

st sumM

j1ψijE

Offij leE

liforalliisinN (11)

sumM

j1ψijT

Offij leT

liforalliisinN (12)

sumN

i1ψijP


sumM












R0ij wjlog2 1 +

Ptri Gi

σ21113888 1113889 (16)




longs to Gl where

EOffij

DiPtri

R0ij

+OiP

idi

Cci

(17)


ηi GiP

tri

Eli

1113969 (18)







i Gij andbj EOff










i Eli

radic




Buyer Mobile device

Wireless channel


Bids collection

AuctionAllocation

Pricing

Response

helliphellip





maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in










tli OiCli (1)



li (2)





EOffij


Cloudcontroller

Data center

Internet



MEC server








(3)



i Gi


⎛⎝ ⎞⎠ (4)



ttrij

Di

Rij

(5)



Etrij

Di

Rij

Ptri (6)






tci

Oi

Cci

(7)



Eidi

OiPidi

Cci

(8)






EOffij

DiPtri

Rij

+OiP

idi

Cci

(9)








minψij1113864 1113865

F sumN

i11minusψij1113872 1113873E

li + ψijE

Offij1113874 1113875 (10)

st sumM

j1ψijE

Offij leE

liforalliisinN (11)

sumM

j1ψijT

Offij leT

liforalliisinN (12)

sumN

i1ψijP


sumM












R0ij wjlog2 1 +

Ptri Gi

σ21113888 1113889 (16)




longs to Gl where

EOffij

DiPtri

R0ij

+OiP

idi

Cci

(17)


ηi GiP

tri

Eli

1113969 (18)







i Gij andbj EOff










i Eli

radic




Buyer Mobile device

Wireless channel


Bids collection

AuctionAllocation

Pricing

Response

helliphellip





maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in









(3)



i Gi


⎛⎝ ⎞⎠ (4)



ttrij

Di

Rij

(5)



Etrij

Di

Rij

Ptri (6)






tci

Oi

Cci

(7)



Eidi

OiPidi

Cci

(8)






EOffij

DiPtri

Rij

+OiP

idi

Cci

(9)








minψij1113864 1113865

F sumN

i11minusψij1113872 1113873E

li + ψijE

Offij1113874 1113875 (10)

st sumM

j1ψijE

Offij leE

liforalliisinN (11)

sumM

j1ψijT

Offij leT

liforalliisinN (12)

sumN

i1ψijP


sumM












R0ij wjlog2 1 +

Ptri Gi

σ21113888 1113889 (16)




longs to Gl where

EOffij

DiPtri

R0ij

+OiP

idi

Cci

(17)


ηi GiP

tri

Eli

1113969 (18)







i Gij andbj EOff










i Eli

radic




Buyer Mobile device

Wireless channel


Bids collection

AuctionAllocation

Pricing

Response

helliphellip





maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in




minψij1113864 1113865

F sumN

i11minusψij1113872 1113873E

li + ψijE

Offij1113874 1113875 (10)

st sumM

j1ψijE

Offij leE

liforalliisinN (11)

sumM

j1ψijT

Offij leT

liforalliisinN (12)

sumN

i1ψijP


sumM












R0ij wjlog2 1 +

Ptri Gi

σ21113888 1113889 (16)




longs to Gl where

EOffij

DiPtri

R0ij

+OiP

idi

Cci

(17)


ηi GiP

tri

Eli

1113969 (18)







i Gij andbj EOff










i Eli

radic




Buyer Mobile device

Wireless channel


Bids collection

AuctionAllocation

Pricing

Response

helliphellip





maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in







i Gij andbj EOff










i Eli

radic




Buyer Mobile device

Wireless channel


Bids collection

AuctionAllocation

Pricing

Response

helliphellip





maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in






maxψij1113864 1113865

F sumN

i1Pi minus sum

M

j1sumN




bdj Cj minussum

Nr1rne iψrjP

tri Gi1113872 1113873EOff

ij


tri Gi

1113969 (20)





F sumN

i1Pi minus sum

M

j1sumN



Θ sumM

j1sumN

i1ψijbj (22)



prime Go2 while Go



j

11 if EOffij leEl





prime Goprime i

24 end while






minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in







minusF bj1113874 1113875ge 0 (23)





minusF bj1113874 1113875ge 0 (24)



minusF bi1113872 11138731113876 1113877

FBM bj1113864 1113865


1113874 1113875

(25)
























i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in












i


50 200 500 800 10000

5

10

15

20

25

30

35


Aver

age e

nerg

y co

nsum

ptio

n of

mob

ile d

evic

es (J

)


Variance









102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in










102 1030

1000

2000

3000

4000

5000

6000

7000


Aver

age t

hrou

ghpu

t of m

obile

dev

ices

(bps

)



0

5

10

15

20

25

30

35

Aver

age d

elay

of m

obile

dev

ices

(s)

Variance






5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in





5 Conclusion




Acknowledgments


References

















102 10309991

09992

09993

09994

09995

09996

09997

09998

09999

1


Ener

gy effi

cien

cy fa

ctor


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in


























Advances in

FuzzySystems


Volume 2018













Journal of

Journal of



Hindawi


Advances in

Multimedia


Biomedical Imaging





RoboticsJournal of









Volume 2018



Advances in




anenergy-awaretaskoffloadingmechanisminmultiuser mobile...

Documents