shodhganga.inflibnet.ac.inshodhganga.inflibnet.ac.in/bitstream/10603/74979/11/11_chapter 2.pdf01 2 $...

17
51 CHAPTER 2 PERFORMANCE EVALUATION OF ROUTING ALGORITHMS FOR MOBILITY MODELS IN AD HOC NETWORK In this chapter, the performance of existing routing strategies in ad hoc networks is investigated. Routing protocols such as DSDV, DSR and AODV are chosen and simulated in a common wireless network simulation platform using Network Simulator version 2 (NS2). In addition to the performance study, mobility models are also proposed and implemented, which are more realistic, and the performance of these routing protocols is compared in more realistic scenarios. Related works (Broach et al 1998, Das et al 1998, Jhonson et al 1999 and Lee et al 1999) that also perform comparative evaluation of ad hoc routing protocols can be found in the literature. However, these articles compare the protocols and use only a single mobility models. These papers evaluate a single class of protocols using performance metrics such as throughput and pure control overhead that only show the effectiveness of the protocol. In this chapter, the performance of existing protocols is investigated for different categories in various network scenarios such as differences in mobility models, mobility rates, traffic patterns, etc. The proposed mobility model defines an exponential distribution model for speed and position with its intuitively more appealing natural formulation than earlier assumptions. This model is applied to

Upload: others

Post on 31-Oct-2019

22 views

Category:

Documents


0 download

TRANSCRIPT

51

CHAPTER 2

PERFORMANCE EVALUATION OF ROUTING

ALGORITHMS FOR MOBILITY MODELS IN AD HOC

NETWORK

In this chapter, the performance of existing routing strategies in ad

hoc networks is investigated. Routing protocols such as DSDV, DSR and

AODV are chosen and simulated in a common wireless network simulation

platform using Network Simulator version 2 (NS2). In addition to the

performance study, mobility models are also proposed and implemented, which

are more realistic, and the performance of these routing protocols is compared

in more realistic scenarios.

Related works (Broach et al 1998, Das et al 1998, Jhonson et al 1999

and Lee et al 1999) that also perform comparative evaluation of ad hoc routing

protocols can be found in the literature. However, these articles compare the

protocols and use only a single mobility models. These papers evaluate a single

class of protocols using performance metrics such as throughput and pure

control overhead that only show the effectiveness of the protocol. In this

chapter, the performance of existing protocols is investigated for different

categories in various network scenarios such as differences in mobility models,

mobility rates, traffic patterns, etc. The proposed mobility model defines an

exponential distribution model for speed and position with its intuitively more

appealing natural formulation than earlier assumptions. This model is applied to

52

different routings protocols and its network performance is evaluated with

different mobility patterns. The ultimate purpose of this work is to find which

routing strategy is best for which environment.

2.1 ROUTING PROTOCOLS

Routing protocols proposed for mobile ad hoc networks are

categorized by reactive and proactive protocols. In this chapter, in order to

investigate the performance of the existing protocol for MANET one proactive

and two reactive protocols are chosen and discussed in the following sections.

2.1.1 Destination Sequenced Distance Vector

DSDV is a table driven algorithm based on the classical Bellman-

Ford or Routing Information Protocol (RIP) routing mechanism. The

improvement made to the Bellman-Ford algorithm includes freedom from loops

in routing tables by using the concept of sequence numbers.

Every mobile in the network maintains a routing table in which all of

the possible destinations within the network and the number of hops to each

destination are recorded. Each entry is marked by a sequence number assigned

by the destination node. The sequence numbers enable the mobile nodes to

distinguish stale routes from new ones, thereby avoiding the formation of

routing loops. To maintain the consistency of routing tables in a dynamically

varying topology each node periodically transmits updates immediately when

significant new information is available.

53

The data broadcast by each mobile node for each new route contains

its sequence number and destination address, the number of hops (metric)

required and the sequence number to reach the destination determined by the

destination. The route with the most recent sequence number is always used. In

the event of two updates having the same sequence number, the route with the

smaller metric is used, in order to optimize the path. Mobile nodes cause broken

links as they move from place to place. A broken link is described by a metric

of infinity. When a link to a next hop is broken in any route, it is immediately

assigned an infinity metric and updated with an odd sequence number.

To avoid periodic updates, which generate a large amount of network

traffic, DSDV supports full dump and incremental updates. In the first case, the

routing packets carry all available information and require multiple Network

Protocol Data Units. Whereas in the second case smaller incremental packets

are used to relay only that information which has changed since the last full

dump. Each of these broadcasts should fit into a standard size NPDU, thereby

decreasing the amount of traffic generated. In DSDV, the mobiles also keep

track of the setting time of routes and, by delaying the broadcast of a routing

update, by the length of the setting time, the network traffic is reduced.

2.1.2 Dynamic source routing

The DSR is an on-demand routing protocol based on the concept of

source routing. This protocol allows nodes to dynamically discover a source

route across multiple network hops to any destination in the ad hoc network.

When using source routing, each packet to be routed carries in its header the

complete ordered list of nodes through which the packet must pass.

54

The protocol consists of two major phases viz Route discovery and Route

maintenance.

When a mobile node wants to send a packet to some destination it

checks its route cache and whether it has any route to the destination. If it has

an unexpired route, it will use this route to send a packet to the destination.

Otherwise, it will initiate a route discovery procedure by broadcasting a RREQ.

Each node hears the route request packet and adds its own address to a source

route. Node S generates a route request to node D and broadcasts the route

request. Nodes B, C and E in turn receive a RREQ from S. Suppose node C

receives the RREQ for the first time it add its address in the route request

packet and rebroadcasts it to the next node H and ignores it if it is already

received. The forwarding of the route request is constructed so that copies of

the request are propagated hop-by- hop outward from the source node until

either the target of the request is found or until another node is found that can

supply a route to the target.

Route reply is generated when the route request reaches either the

destination itself or an intermediate node, which contains in its route cache an

unexpired route to the destination. Figure 2.1 shows that destination D, on

receiving the first RREQ, sends a RREP on a route obtained by reversing the

route appended to receive the RREQ. The RREP includes the route from S to D

on which the RREQ was received by node D. If the node generating the route

reply is the destination, it places the route record contained in the route request

into the route reply. If the responding node is an intermediate node it appends

its cached route to the route record and then generates the route reply.

55

RREP [S,C,H,D]

Figure 2.1 Route Reply in DSR

Route maintenance is the process of monitoring the status of a source

route while in use, so that any link-failure along the source route can be

detected and the broken link removed from use. Route error packets and

acknowledgements are used to maintain the route. Route error packets are

generated at a node when the data link layer encounters a fatal transmission

problem. When a route error packet is received, the hop in error is removed

from the node route cache and all routes containing the hop are truncated at that

point. In addition route error message acknowledgements are used to verify the

correct operation of the route links. The parameter values of DSR used for the

simulation are given in Table 2.1.

Table 2.1 Parameter values for DSR

Time between retransmitted Route Requests 500 msecs

Max. time where the same request can be sent 10 secs

2.1.3 Ad hoc On-Demand Distance Vector

AODV is an improvement on the DSDV algorithm because it

minimizes the number of required broadcasts by creating routes on demand

56

basis as opposed to maintaining a complete list of routes in the DSDV

algorithms. In the AODV algorithm the flooded route request packets are used

to find routes as done in DSR. The AODV algorithm maintains a routing table

in all intermediate nodes instead of a route cache. Nodes forwarding the request

remember the earlier hop taken by the request packet. This hop information is

used to forward the reply packet back to the source. The route reply packet sets

up the routing table entries on its path. If any node has a route to the required

destination, it can reply to a request. It also uses a technique called route expiry,

where a routing table entry expires after a predetermined period, after which

fresh route discovery must be initiated. It uses HELLO messages to determine

the connectivity among the neighbors.

2.2 MOBILITY MODELS

The mobility models in ad hoc networks deal with individual motion

behavior of the nodes. Many researchers use the random mobility model

(Zonoozi and Dassznayake 1997) according to which the speed and direction of

motion in a new time interval have no relation to their past values. The

modified version of the random mobility model (Basagni et al 1998) is used in

DREAM protocol. The mobile host has a random direction at every simulation

clock tick, but a constant speed during the entire simulation period. In Ko’s

simulation model (Ko and Vaidya, 1998) the mobile hosts are allowed to move

along a path that is made up of several segments that are exponentially

distributed. The direction of each segment is chosen in a random manner.

According to Das et al model (1998), a node chooses its speed,

direction and distance based on predefined uniform distribution, and then

calculates its next destination and the time to reach the destination. Johnson’s

57

model (1996) is an extension of the random walk. Here the mobile host first

stays at a location for a certain time and then moves to a new random chosen

destination at a speed uniformly distributed between [0, MaxSpeed]. Haas

(1997) presents an incremental model in which the speed and direction of

current movement randomly diverge from the previous speed and direction after

each time increment. However, Sanchez (1998) studies the relationship among

mobile hosts that move with the same purpose. For example, in a military

environment, it is most likely that several mobile hosts move with a common

objective.

2.3 PERFORMANCE EVALUATION

The performance of the mobile ad hoc network for various mobility

models is studied with the help of the Network Simulator package NS2 (Kevin

Fall and Kannan Varadhan 1998). In NS2 the initial network topology is

generated by the random position of the node. Upon receiving the various

simulation input parameters such as the number of nodes (N), simulation area

(A), speed (s), pause time (tp) and simulation time (T), mobility is initiated. The

node positions are then updated and the shortest path and number of hops

required to reach the destination are computed. Based on the number of active

nodes (me) and their respective data rates the packets are routed to the required

destination. The trace file containing the status of the nodes (receive, send) and

links (failure, connectivity), types of packets (control, data, acknowledgment)

and average hop count is then generated and the process goes on till the

simulation time is over. The performance parameters such as throughput,

overhead and delay of the simulated mobile network are then computed. These

parameters are defined as

58

Throughput: Measured as the ratio of the no. of data packets delivered to the

destination and the no. of data packets sent by the sender. This number presents

the effectiveness of the protocol.

End-to-end delay : Measured in ms as the time between the reception of the

last and first packet / total no. of packets reaching the application layer. This

delay includes processing and queuing delays in each intermediate node.

Control overhead : Measured as the ratio of the number of control packets

transmitted during the entire simulation period by data packets transmitted.

The simulation procedure in NS2 can be described as shown in

Figure 2.2. The network simulator assumes the mobile speed as a uniformly

distributed random process. In this study, more realistic models with

exponential and normal distribution, which assign lower probabilities for higher

speed are considered. The simulation of the mobile network is carried out on

800MHz Pentium III processor, 40GB Hard Disk capacity and Red Hat Linux

version 6.2 operating system with the parameter specifications shown in

Table 2.2.

Table 2.2 Parameters used during simulation

Transmitter range 250mSimulation area 1500m X 300mSimulation time 900 secsNumber of nodes 50Bandwidth 2 MHzTraffic type Constant bit ratePacket size 512

59

Initial Topology

Scenario generation___i___Pause time(tp) simulation

| | time (T)seed'^

speech-^, Initiate mobility

ANode Position

Update

no. of nodes (N) simu /ation /

area (A) / Default mobility/ model: Uniform

Mobility models P(x): exponential/

normal

IComputation of shortest path & no. of hops

Connection pattern

Active node selection

no. of active node selection (me) \no. ofpackets/sec/

source (rate)

v yes

Compute throughput, overhead & dela

no

Figure 2.2 Sequence of simulation using NS2

60

The performance of Dynamic Source Routing for various mobility

models is plotted in Figures 2.3, 2.4 and 2.5. The throughput performance of the

DSR algorithm shown in Figure 2.3 indicates that the exponential model gives a

steady state throughput of 98% whereas in the uniform and normal distribution

models it progressively decreases to 95% and 90% respectively as the speed is

varied from lm/s to 5m/s. This could possibly be due to the fact that the dwell-

in time probability of the mobiles at higher speeds is small on the exponential

model than in the uniform and normal distribution models.

DSR Throughput Performance

Figure 2.3 DSR Throughput comparisons for various models

The end-to-end delay is very small at lower speeds upto 2m/s

whereas at higher speeds it increases marginally in the uniform and exponential

models but rapidly in the for normal model as shown in Figure 2.4. This might

be due to link breakages at higher speeds and as a consequence more number of

hops are required to reach the destination. At very low speeds, the control

overhead in the normal mobility model is found to be much lower than in other

61

models, which could be because of smaller lower end tail probability and is

shown in Figure 2.5. But, at higher speeds the effect of high-end tail probability

for the normal model is observed in the decreasing trend of the overhead. There

is approximately a linear change in control overhead with speed in the uniform

model whereas it is almost constant in the exponential model.DSR Delay Performance

0.7 ,--------------r--------------- ,---------------- ,---------------- ,---------------- r- i * i ----------j

-q Uniform -e— Exponential

0.6 - —Gaussian

1 1 5 2 2.5 3 3.5 4 4.5 5Speed (m/s)

Figure 2.4 DSR Delay comparisons for various models

1 1-5 2 2.5 3 3.5 4 4.5 5S peed {m/s)

Figure 2.5 DSR Overhead comparisons for various models

62

The results of AODV algorithm for various mobility models are

shown in Figures 2.6, 2.7 and 2.8. At the highest mobility, AODV produces less

delay compared to DSR and also delivers more number of packets than the

other two algorithms with minimum packet delivery ratio of 93% for Gaussian

models. The overhead shown in Figure 2.7 clearly exposes those characteristics

of the model. The Exponential model experiences less overhead at higher

mobility than the other two models due to lower probability mobility at higher

speed. When compared to the DSR algorithm the number of control packets are

more for AODV, since it uses HELLO packets periodically, resulting in higher

packet overhead.AODV Throughput Performance

Figure 2.6 AODV Throughput comparisons for various models

Speed (m/s)

Figure 2.7 AODV Delay comparisons for various models

63

AODV Overhead Performance

Figure 2.8 AODV Overhead comparisons for various models

The performance of DSDV shown in Figures 2.9, 2.10 and 2.11

indicates that there is slightly lower throughput at lower mobility compared to

DSR and AODV. This is due to the fact that the packets are sent before routes

converge in the network. As mobility increases more number of packets are

dropped due to periodic route update in DSDV. The delay performance of

DSDV shown in Figure 2.10 exhibits less delay than the DSR for normal

models at high speed because only the packets belonging to valid routes at the

sending instant get through. But, more number of packets is lost until new route

entries are propagated through the network by the route updates. The overhead

performance of DSDV, which is shown in Figure 2.11 for all the mobility

models is very high when compared to those of DSR and AODV.

# of c

ontro

l pac

kets

64

DSDV Throughput Performance

Figure 2.9 DSDV Throughput comparisons for various models

DSDV Delay Performance

Speed (m/s)

Figure 2.10 DSDV Delay comparisons for various models

65

DSDV Overhead Performance

Figure 2.11 DSDV Overhead comparisons for various models

The comparative performance of all three routing algorithms for

various mobility models is given in Table 2.3 and Table 2.4 gives the parameter

ranges for all three algorithms.

Table 2.3 Comparative performance of ail three routing algorithms for

various mobility models

\ Mobility \models

Routing \ Algorithms \

% of Throughput Overhead Delay (secs)

UniformExpo­nential Gaussian

Uniform Expo­nential

GaussianUniform Expo­

nentialGaussian

DSR High High Low Low Low Medium Low Low Medium

AODV High High Medium Medium Medium Low Medium Low High

DSDV Medium Medium Low Very high Very high Very high Medium Low High

66

Table 2.4 Parameter ranges

RangesParameter\^^ Low Medium High

VeryHigh

Throughput 85 - 90 % 90-95% 95-100 % -

Delay < 0.05 sec 0.05-0.1 >0.1 -

Overhead <2000 2000 - 5000 5000-10000 >10000

The numerical result of the relative performance of the routing

algorithm for various traffic load conditions and the rate of transmission of

packets are given in Tables 2.5 and 2.6. For DSDV algorithm with Gaussian

mobility model the network simulator could generate a scenario and define the

connection pattern but does not generate the trace file and therefore the results

are not given in Table 2.6.

Table 2.5 Parameter comparison for various mobility models under

various traffic loads

Speed = 3m/s # of packets per sec = 4

RoutingAlgorithms

TrafficLoad(me)

% of Throughput Overhead Delay (secs)

UniformExpo­nential Gaussian

Uniform Expo­nential

GaussianUniform Expo­

nentialGaussian

DSR

5 95.14 97.26 91.08 861 908 981 .04154 .0714 .127610 97.38 96.91 90.19 1120 1551 1893 .02365 .0367 .353715 97.2 97.36 91.59 2179 1865 4615 .04329 .0517 .161820 1 97.38 92.1 3091 2939 4561 .02524 .03645 .063830 97.1 97.37 91.27 4995 3567 6340 .07231 .02574 .2805

AODV

5 95.70 95.98 93.36 4195 5241 5943 .01166 .01577 .022210 96.42 96.32 93.98 6549 5557 7073 .02055 .02728 .022715 96.63 97.15 94.69 9685 9445 14500 .01152 .01194 .022120 96.34 95.97 94.48 13.8*10’ 12.2*10’ 1.91*10’ .01245 .01159 .012930 96.56 96.78 90.02 18.9*10’ 16.1*10’ 24.1*10’ .01234 .01347 .0861

DSDV

5 92.44 94.65 93.31 44.9*10’ 44.7*10’ 43.2*10’ .07925 .04150 .058310 93.93 91.38 90.15 44.8*10’ 40.8*10’ 43.7*10’ .08212 .02375 .05715 94.20 90.34 90.03 44.9*10’ 44.7*10’ 43.6*10’ .05067 .02346 .081520 2.78 93.23 8822 44.4*10’ 40.2*10’ 44.4*10’ .02172 .03645 .058330 94.23 93.07 89.22 44.8*10’ 44.6*10’ 45.0*10’ .07349 .03255 .2837

67

Table 2.6 Parameter comparison for various mobility models under

different speeds

me = 10 # of packets per see = 10

RoutingAlgorithms

Speed(m/sec)

% of Throughput Overhead Delay (secs)

UniformExpo­nential Gaussian Uniform

Expo­nential

GaussianUniform Expo­

nentialGaussian

DSR

i 100 95.31 99.24 8 1083 67 .0029 .0903 .01242 93.33 92.93 89.13 1285 1359 1196 .0556 .1572 .04243 93.28 92.69 84/72 1370 1493 1113 .0502 .1059 .09774 66.24 92.39 81.85 1947 1308 2490 .2549 .2148 .12185 90.59 100 81.92 1214 8 2684 .1872 .0029 .4315

AODV

1 96.70 94.68 96.99 3819 3385 2837 .0236 .0517 .01752 93.73 93.73 95.32 5541 4944 4037 .0637 .0158 .04783 92.61 92.85 88.72 5528 5140 10.6*103 .0368 .0292 .02824 93.1 93.52 86.55 6407 6099 8.3*103 .2369 .0214 .01395 91,94 92.77 84.14 7388 5757 10.9*103 .1983 .6813 .4060

DSDV

1 97.12 96.39 44.9*103 44.9*103 .0131 .01392 94.54 94.52 44.8*103 44.8*103 .0122 .04883 92.32 91.86 44.9*103 44.6*103 .0283 .01374 91.17 92.65 48.9* 103 44.9*103 .0546 .01645 93.45 91.95 44.8*103 44.5*103 .0494 .0366

2.4 SUMMARY

In this chapter, the effect of various mobility models in ad hoc

networks with DSR, AODV and DSDV routing algorithms are studied. The

performance results of the all three algorithm are presented in detail. From the

simulation, it is observed that AODV and DSDV requires more control over

head as the network traffic increases. DSR is more suitable for smaller network

due to source routing. Also from the results of the throughput, end-to-end delay

and the control overhead parameters, it is found that the exponential model

outperforms the uniform and normal models for a wide range of speeds and

traffic loads. Based on this study, it is concluded that the exponential mobility

models are more suitable for the analysis and simulation of mobile ad hoc

networks.