IACSIT International Journal of Engineering and Technology, Vol.2, No.

2, April 2010
ISSN: 1793-8236

Performance analysis of AODV, DSR & TORA

Routing Protocols
Anuj K. Gupta, Member, IACSIT, Dr. Harsh Sadawarti, Dr. Anil K. Verma

has been paid to use specific network parameters when

Abstract- The field of Mobile Ad hoc Networks (MANETs) specifying routing metrics. Examples might include delay of
has gained an important part of the interest of researchers and the network, link capacity, link stability or identifying low
become very popular in last few years. MANETs can operate
without fixed infrastructure and can survive rapid changes in mobility nodes. These schemes are generally based on
the network topology. They can be studied formally as graphs previous work, which is then enhanced with the new metrics.
in which the set of edges varies in time. The main method for Paper Outline
evaluating the performance of MANETs is simulation. This
paper is subjected to the on-demand routing protocols with The rest of the paper is organized as follows: Section II
identical loads and environment conditions and evaluates their presents the definition of MANET and its topology. Section
relative performance with respect to the two performance III presents the mobile ad hoc routing protocols categories.
metrics: average End-to-End delay and packet delivery ratio. Section IV provides an overview and general comparison of
We investigated various simulation scenarios with varying
the routing protocols used in the study. The simulation
pause times. From the detailed simulation results and analysis,
a suitable routing protocol can be chosen for a specified environment and performance metrics are described in
network and goal. Section V and then the results are presented in Section VI.
Finally Section VII concludes the paper.
Index Terms- MANET, AODV, DSR, TORA

II. MOBILE AD HOC NETWORK

I. INTRODUCTION A MANET topology can also be defined as a dynamic
The history of wireless networks started in the 1970s and (arbitrary) multi-hop graph G = (N, L), where N is a finite
the interest has been growing ever since. At present, this set of mobile nodes (MNs) and L is a set of edges which
sharing of information is difficult, as the users need to represent wireless links. A link (i, j) Є L exists if and only if
perform administrative tasks and set up static, bi-directional the distance between two mobile nodes is less or equal than
links between the computers. This motivates the a fixed radius r as shown. This r represents the radio
construction of temporary networks with no wires, no transmission range that depends on wireless channel
communication infrastructure and no administrative characteristics including transmission power. Accordingly,
intervention required. Such interconnection between mobile the neighborhood of a node x is defined by the set of nodes
computers is called an Ad hoc Network. Ad hoc networks that are inside a circle (assume that MNs are moving in a
are emerging as the next generation of networks and defined two-dimensional plane) with center at x and radius r, and it
as a collection of mobile nodes forming a temporary is denoted by:
(spontaneous) network without the aid of any centralized
administration or standard support services. In Latin, ad hoc N r ( x) = N x = {n f | d ( x, n f } ≤ r , x ≠ n f , ∀j ∈ N , j ≤| N |}
literally means “for this,” further meaning “for this purpose
where x is an arbitrary node in graph G and d is a distance
only” and thus usually temporary [1]. An ad hoc network is
function [8].
usually thought of as a network with nodes that are
relatively mobile compared to a wired network. Hence the
A path (route) from node i to node j, denoted by Rij is a
topology of the network is much more dynamic and the
sequence of nodes Rij=(i,n1, n2,…,nk, j) where (i,n1), (nk,j)
changes are often unpredictable oppose to the Internet
and (ny,ny+1) for 1≤ y ≤ k-1 are links. A simple path from i to
which is a wired network. This fact creates many
j is a sequence of nodes with no node being repeated more
challenging research issues, since the objectives of how
than once. Due to the mobility of the nodes, the set of paths
routing should take place is often unclear because of the
(wireless links) between any pair of nodes and distances is
different resources like bandwidth, battery power and
changing over time. New links can be established and
demands like latency.
existing links can vanish.
MANETs have several salient characteristics: 1)
Dynamic topologies 2) Bandwidth constrained, variable
capacity links 3) Energy-constrained operation 4) Limited III. ROUTING PROTOCOLS
physical security. Therefore the routing protocols used in Routing protocols for Mobile ad hoc networks can be
ordinary wired networks are not well suited for this kind of broadly classified into two main categories:
dynamic environment. Routing algorithms are often
difficult to be formalized into mathematics they are instead • Proactive or table-driven routing protocols
tested using extensive simulation. Recently more attention • Reactive or on-demand routing protocols.
226
 IACSIT International Journal of Engineering and Technology, Vol.2, No.2, April 2010
ISSN: 1793-8236
A. Table Driven Routing Protocols (Proactive) source to destination, it does not add any overhead to the
In proactive or table-driven routing protocols, each node packets. However, route discovery process is only initiated
continuously maintains up-to-date routes to every other when routes are not used and/or they expired and
node in the network. Routing information is periodically consequently discarded. This strategy reduces the effects of
transmitted throughout the network in order to maintain stale routes as well as the need for route maintenance for
routing table consistency. Thus, if a route has already unused routes. Another distinguishing feature of AODV is
existed before traffic arrives, transmission occurs without the ability to provide unicast, multicast and broadcast
delay. Otherwise, traffic packets should wait in queue until communication. AODV uses a broadcast route discovery
the node receives routing information corresponding to its algorithm and then the unicast route reply massage.
destination. However, for highly dynamic network topology, B. Dynamic Source Routing (DSR)
the proactive schemes require a significant amount of
The Dynamic Source Routing (DSR) [6] is one of the
resources to keep routing information up-to-date and
purest examples of an on-demand routing protocol that is
reliable. Certain proactive routing protocols are Destination-
based on the concept of source routing. It is designed
Sequenced Distance Vector (DSDV), Wireless Routing
specially for use in multihop ad hoc networks of mobile
Protocol (WRP), Global State Routing (GSR) and Cluster-
nodes. It allows the network to be completely self-
head Gateway Switch Routing (CGSR).
organizing and self-configuring and does not need any
B. On-Demand Routing Protocols (Reactive) existing network infrastructure or administration. DSR uses
In contrast to proactive approach, in reactive or on no periodic routing messages like AODV, thereby reduces
demand protocols, a node initiates a route discovery network bandwidth overhead, conserves battery power and
throughout the network, only when it wants to send packets avoids large routing updates. Instead DSR needs support
to its destination. For this purpose, a node initiates a route from the MAC layer to identify link failure. DSR is
discovery process through the network. This process is composed of the two mechanisms of Route Discovery and
completed once a route is determined or all possible Route Maintenance, which work together to allow nodes to
permutations have been examined. Once a route has been discover and maintain source routes to arbitrary destinations
established, it is maintained by a route maintenance process in the network. DSR has a unique advantage by virtue of
until either the destination becomes inaccessible along source routing. As the route is part of the packet itself,
every path from the source or until the route is no longer routing loops, either short – lived or long – lived, cannot be
desired. In reactive schemes, nodes maintain the routes to formed as they can be immediately detected and eliminated.
active destinations. A route search is needed for every This property opens up the protocol to a variety of useful
unknown destination. Therefore, theoretically the optimizations.
communication overhead is reduced at expense of delay due Neither AODV nor DSR guarantees shortest path. If the
to route research. Some reactive protocols are Cluster Based destination alone can respond to route requests and the
Routing Protocol (CBRP), Ad hoc On-Demand Distance source node is always the initiator of the route request, the
Vector (AODV), Dynamic Source Routing (DSR), initial route may the shortest.
Temporally Ordered Routing Algorithm (TORA), C. Temporary Ordered Routing Algorithm (TORA)
Associativity-Based Routing (ABR), Signal Stability
The Temporally Ordered Routing Algorithm (TORA) is a
Routing (SSR) and Location Aided Routing (LAR).
highly adaptive, efficient and scalable distributed routing
algorithm based on the concept of link reversal [3]. TORA
IV. OVERVIEW OF AODV, DSR AND TORA is proposed for highly dynamic mobile, multi-hop wireless
networks. It is a source-initiated on-demand routing
Every routing protocol has its own merits and demerits,
protocol. It finds multiple routes from a source node to a
none of them can be claimed as absolutely better than others.
destination node. The main feature of TORA is that the
We have selected the three reactive routing protocols –
control messages are localized to a very small set of nodes
AODV, DSR and TORA for evaluation [11,18].
near the occurrence of a topological change. To achieve this,
A. Ad hoc On-demand Distance Vector Routing (AODV) the nodes maintain routing information about adjacent
Ad-hoc On-demand distance vector (AODV) [4,16] is nodes. The protocol has three basic functions: Route
another variant of classical distance vector routing creation, Route maintenance and Route erasure. TORA can
algorithm, a confluence of both DSDV [5] and DSR [6]. It suffer from unbounded worst-case convergence time for
shares DSR’s on-demand characteristics hence discovers very stressful scenarios [15,17]. TORA has a unique feature
routes whenever it is needed via a similar route discovery of maintaining multiple routes to the destination so that
process. However, AODV adopts traditional routing tables; topological changes do not require any reaction at all. The
one entry per destination which is in contrast to DSR that protocol reacts only when all routes to the destination are
maintains multiple route cache entries for each destination. lost. In the event of network partitions the protocol is able to
The initial design of AODV is undertaken after the detect the partition and erase all invalid routes.
experience with DSDV routing algorithm. Like DSDV, Table 1 lists some comparisons between the three routing
AODV provides loop free routes while repairing link protocols discussed above.
breakages but unlike DSDV, it doesn’t require global
periodic routing advertisements. AODV also has other
significant features. Whenever a route is available from

227
 IACSIT International Journal of Engineering and Technology, Vol.2, No.2, April 2010
ISSN: 1793-8236
TABLE 1.COMPARISON OF THE THREE ROUTING PROTOCOLS at the interface queue, retransmission delays at the MAC,
Parameters AODV DSR TORA and propagation and transfer times. It can be defined as:
Source No Yes No
s
1
∑ (r − s )
Routing
Topology Full Full Reduced D= i i
Broadcast Full Full Local N i =1
Update Route Route Node’s where N is the number of successfully received packets, i is
information error error height unique packet identifier, ri is time at which a packet with
Update Source Source Neighbors
destination unique id i is received, si is time at which a packet with
Method Unicast Unicast Broadcast unique id i is sent and D is measured in ms. It should be less
Storage O(E) O(E) O(Dd*A) for high performance.
Complexity
Abbreviations:
Dd – Number of maximum desired destinations VI. SIMULATION RESULTS & OBSERVATIONS
E – Communication pairs
A – Average number of adjacent nodes The simulation results are shown in the following section
in the form of line graphs. Graphs show comparison
V. SIMULATION between the three protocols by varying different numbers of
sources on the basis of the above-mentioned metrics as a
The simulations were performed using Network function of pause time.
Simulator 2 (Ns-2) [2], particularly popular in the ad hoc
networking community. The traffic sources are CBR A. Packet Delivery Fraction (PDF) or Throughput
(continuous bit –rate). The source-destination pairs are Figure 1 a-c, shows a comparison between the routing
spread randomly over the network. protocols on the basis of packet delivery fraction as a
The mobility model uses ‘random waypoint model’ in a function of pause time and using different number of traffic
rectangular filed of 500m x 500m with 50 nodes. During the sources. Throughput describes the loss rate as seen by the
simulation, each node starts its journey from a random spot transport layer. It reflects the completeness and accuracy of
to a random chosen destination. Once the destination is the routing protocol. From these graphs it is clear that
reached, the node takes a rest period of time in second and throughput decrease with increase in mobility. As the
another random destination is chosen after that pause time. packet drop at such a high load traffic is much high.
This process repeats throughout the simulation, causing TORA performs better at high mobility but in other cases
continuous changes in the topology of the underlying it shows to have a lower throughput. AODV in our
network. Different network scenario for different number of simulation experiment shows to have the best overall
nodes and pause times are generated. The model parameters performance. On-demand protocols (DSR and AODV) drop
that have been used in the following experiments are a considerable number of packets during the route discovery
summarized in Table 2. phase, as route acquisition takes time proportional to the
distance between the source and destination. The situation is
TABLE 2. SIMULATION PARAMETERS similar with TORA. Packet drops are fewer with proactive
Parameter Value
Simulator ns-2
protocols as alternate routing table entries can always be
Protocols studied AODV, DSR and TORA assigned in response to link failures. TORA can be quite
Simulation time 200 sec sensitive to the loss of routing packets compared to the
Simulation area 500 x 500 other protocols. Buffering of data packets while route
Transmission range 250 m
Node movement model Random waypoint
discovery in progress, has a great potential of improving
Bandwidth 2 MBit DSR, AODV and TORA performances. AODV has a
Traffic type CBR (UDP) slightly lower packet delivery performance than DSR
Data payload Bytes/packet because of higher drop rates. AODV uses route expiry,
Bandwidth 2 Mbps
dropping some packets when a route expires and a new
Performance Indices route must be found [13].
The following performance metrics are considered for
evaluation:
100
98
Packet Delivery Ratio (%)

Packet Delivery Fraction (PDF): The ratio of the data

96
packets delivered to the destinations to those generated by
94
the sources. Mathematically, it can be expressed as:
92
1 e Rf
P= ∑ 90

c f =1 N f 88
86
where P is the fraction of successfully delivered packets, C
84
is the total number of flow or connections, f is the unique 0 10 20 50 100 150 200 250 300
flow id serving as index, Rf is the count of packets received
Paus e Tim e (s e c)
from flow f and Nf is the count of packets transmitted to f.
AODV DSR TORA
Average end-to-end delay: This includes all possible delays
caused by buffering during route discovery latency, queuing (a)

228
 IACSIT International Journal of Engineering and Technology, Vol.2, No.2, April 2010
ISSN: 1793-8236

100 18
16

End To End Delay (sec)

98
Packet Delivery Ration (%)

14
96 12
94 10
92 8
90 6
4
88
2
86 0
84 0 20 50 100 150 200 250 300 400 500
82 Paus e Tim e (s e c)
80
0 20 50 100 150 200 250 300 400 500 DSR A ODV TORA

Paus e Tim e (s e c) (c)

AODV DSR TORA Figure 2. End to End Delay vs. Pause time for 50-node model with (a) 10
sources, (b) 20 sources and (c) 50 sources.
(b)
B. End to End Delay
120 Figure 2 a-c, shows the graphs for end-to-end delay Vs
100 pause time. From these graphs we see that the average
Packet Delivery Ratio (%)

packet delay increase for increase in number of nodes

80
waiting in the interface queue while routing protocols try to
60 find valid route to the destination. Besides the actual
delivery of data packets, the delay time is also affected by
40
route discovery, which is the first step to begin a
20 communication session. The source routing protocols have a
0
longer delay because their route discovery takes more time
0 20 50 100 150 200 250 300 350 400 450 500 as every intermediate node tries to extract information
Paus e Tim e (s e c)
before forwarding the reply. The same thing happens when
a data packet is forwarded hop by hop. Hence, while source
A ODV DSR TORA routing makes route discovery more profitable, it slows
down the transmission of packets.
(c)
Figure 1. Packet delivery fraction vs. Pause time for 50-node model with AODV and DSR show poor delay characteristics as their
(a) 10 sources, (b) 20 sources and (c) 50 sources. routes are typically not the shortest. Even if the initial route
discovery phase finds the shortest route (it typically will),
the route may not remain the shortest over a period of time
25
due to node mobility. However, AODV performs a little
20 better delay-wise and can possibly do even better with some
End to End Delay (sec)

fine-tuning of this timeout period by making it a function of

15
node mobility. TORA too has the worst delay
10 characteristics because of the loss of distance information
with progress. Also in TORA route construction may not
5
occur quickly. This leads to potential lengthy delays while
0 waiting for new routes to be determined. In DSR Route
0 20 50 100 150 200 250 300 400 500
Discovery is fast, therefore shows a better delay
Paus e Tim e (s e c)
performance than the other reactive protocols at low pause
A ODV TORA DSR time (high mobility). But in case of congestion (high traffic)
DSR control messages get loss thus eliminating its
(a)
advantage of fast establishing new route. Under such
situations DSR has a relatively high delay that AODV, but
20
however the delay decreases with increase in pause time
[11].
End To End Delay (sec)

15
Without any periodic hello messages, DSR outperforms
10 the other protocols in terms of overhead. In most cases, both
the packet overhead and the byte overhead of DSR are less
5
than a quarter of AODV’s overhead. AODV has the largest
0 routing load (in the 50-node cases, as many as 6.5 routing
0 20 50 100 150 200 250 300 400 500
packets per data packet and 2 routing bytes per data byte)
Paus e Tim e (s e c)
because the number of its route discoveries is the most, and
DSR A ODV TORA the discovery is network-wide flooding. When there are
(b) more connections, more routing is needed, and so the
proportion of hello messages in the total overhead becomes
smaller. As the result, AODV gets closer to DSR. The
229
 IACSIT International Journal of Engineering and Technology, Vol.2, No.2, April 2010
ISSN: 1793-8236
excellent routing load performance of DSR is due to the performance of three widely used ad hoc network routing
optimizations possible by virtue of source routing. TORA’s protocols using packet-level simulation. The simulation
performance is not very competitive with the distance characteristics used in this research, that is, packet delivery
vector and on-demand protocols. We conjecture that it is fraction and end-to-end delay are unique in nature, and are
due to the fact network partitions cause TORA to do very important for detailed performance evaluation of any
substantial work to erase routes even when those routes are networking protocol.
not in use [13]. However, TORA shows a better We can summarize our final conclusion from our
performance for large networks with low mobility rate. experimental results as follows:
Comparison Study • Increase in the density of nodes yields to an increase in
the mean End-to-End delay.
The goal of this performance evaluation is a comparison • Increase in the pause time leads to a decrease in the
of a MANET between AODV, DSR and TORA routing mean End-to-End delay.
protocols. AODV in our simulation experiment shows to • Increase in the number of nodes will cause increase in
have the overall best performance. It has an improvement of the mean time for loop detection.
DSR and DSDV and has advantages of both of them. In short, AODV has the best all round performance. DSR
TORA performs better at high speed high mobility and has is suitable for networks with moderate mobility rate. It has
a high throughput as compared to AODV and DSR. It often low overhead that makes it suitable for low bandwidth and
serves as the underlying protocol for lightweight adaptive low power network. Whereas TORA is suitable for
multicast algorithms. Whereas DSR suits for network in operation in large mobile networks having dense population
which mobiles move at moderate speed. It has a significant of nodes. The major benefit is its excellent support for
overhead as the packet size is large carrying full routing multiple routes and multicasting.
information.
Table 3 shows a numerical comparison of the three FUTURE WORK
protocols, “1” for the best up to “4” for the worst [14].
In the future, extensive complex simulations could be
TABLE 3. NUMERICAL COMPARISON OF THE THREE ROUTING PROTOCOLS carried out using other existing performance metrics, in
order to gain a more in-depth performance analysis of the ad
Metrics AODV DSR TORA hoc routing protocols. Other new protocols performance
Scalability 2 3 1
Delay 3 2 4
could be studied too.
Routing
2 1 3
overhead ACKNOWLEDGEMENT
Drop packet 1 2 3
Throughput 1 2 4 The authors wish to thank the reviewers and editors for
Dynamic their valuable suggestions and expert comments that help
2 3 1
adaptability improve the paper.
Energy
2 1 3
conservation
REFERENCES
[1] Mobile ad hoc networks (MANET). http: //www.ietf.org/
html.charters/manet-charter.html, 1997. IETF Working Group
VII. CONCLUSIONS Charter
[2] NS-2 Network simulator http://www.isi.edu/nsnam/ns.
As a special type of network, Mobile Ad hoc Networks [3] Vincent D. Park and M.Scott Corson. A highly adaptive distributed
(MANETs) have received increasing research attention in routing algorithm for mobile wireless networks. In Proceedings of
recent years. There are many active research projects INFOCOM 1997, 1997.
[4] C. E. Perkins and E. M. Royer, “Ad Hoc On-demand Distance Vector
concerned with MANETs. Mobile ad hoc networks are Routing,” In Proceedings of the 2nd IEEE Workshop on Mobile
wireless networks that use multi-hop routing instead of Computing Systems and Applications, New Orleans, LA, February
static networks infrastructure to provide network 1999, pp. 90-100.
[5] C. E. Perkins and P. Bhagwat, “Highly Dynamic Destination-
connectivity. MANETs have applications in rapidly Sequenced Distance-Vector Routing (DSDV) for Mobile
deployed and dynamic military and civilian systems. The Computers,” SIGCOMM, London, UK, August 1994, pp. 234-244.
network topology in MANETs usually changes with time. [6] D. B. Johnson and D. A. Maltz, “Dynamic Source Routing in Ad-Hoc
Ad hoc Networks," Mobile Computing, ed. T. Imielinski and H.
Therefore, there are new challenges for routing protocols in Korth, Kluwer Academic Publishers, 1996, pp. 153-181.
MANETs since traditional routing protocols may not be [7] Charles E. Perkins, Ad Hoc Networking, Addison-Wesley, March
suitable for MANETs. Researchers are designing new 2005.
[8] N. Nikaein, H. Labiod, and C. Bonnet. DDR-Distributed Dynamic
MANETs routing protocols, comparing and improving Routing Algorithm for Mobile Ad hoc Networks. First Annual
existing MANETs routing protocols before any routing Workshop on Mobile and Ad Hoc Networking and Computing
protocols are standardized using simulations. (MobiHOC), 2000.
This work is an attempt towards a comprehensive [9] S. R. Das, R. Castaneda, J. Yan, and R. Sengupta, “Comparative
performance evaluation of routing protocols for mobile, ad hoc
performance evaluation of three commonly used mobile ad networks,” in Proceedings of 7th International Conference on
hoc routing protocols (DSR, TORA and AODV). Over the Computer Communications and Networks (IC3N ’98) pp. 153 161,
past few years, new standards have been introduced to Lafayette, La, USA, October 1998.
[10] M. Särelä. Measuring the Effects of Mobility on Reactive Ad Hoc
enhance the capabilities of ad hoc routing protocols. As a Routing Protocols. Helsinki University of Technology Laboratory for
result, ad hoc networking has been receiving much attention Theoretical Computer Science, Research Reports 91, 2004.
from the wireless research community. In this paper, using
the latest simulation environment NS 2, we evaluated the
230
 IACSIT International Journal of Engineering and Technology, Vol.2, No.2, April 2010
ISSN: 1793-8236
[11] Samir R. Das, Charles E. Perkins, Elizabeth E. Royer, “Performance
Comparison of Two On-demand Routing Protocols for Ad Hoc
Networks”.
[12] Johansson, P., T. Larsson, N. Hedman, B. Mielczarek, and M.
Degermark. 1999. Scenario-based performance analysis of routing
protocols for mobile ad-hoc networks. Paper presented at
Mobicom’99, August, Seattle, WA.
[13] Samir R. Das, Robert Castaneda and Jiangtao Yan, “Simulation-
based performance evaluation of routing protocols for mobile ad hoc
networks”.
[14] Samba Sesay, Zongkai Yang, Biao Qi and Jianhua He, “Simulation
comparison of Four Wireless Ad Hoc Routing Protocols”.
[15] Park V. and S. Corson, 2001. Temporary-ordered Routing Algorithm
(TORA). Internet Draft, draft-ietf-manettora-spec-04.txt.
[16] C. Perkins, Ad hoc On demand Distance Vector (AODV) routing,
IETF Internet draft (1997), http://www.ietf.org/internet-drafts/draft-
ietf-manet-aodv-00.txt.
[17] V. Park and S. Corson, Temporally Ordered Routing Algorithm
(TORA) Version 1, Functional specification IETF Internet draft
(1998), http://www.ietf.org/internet-drafts/draft-ietf-manet-tora-spec-
01.txt.
[18] E. M. Royer and C. Toh, “A Review of Current Routing Protocols for
Ad Hoc Mobile Wireless Networks,” IEEE Personal
Communications, pp. 46–55, April 1999.

231