Performance Comparison of AODV, DSR and OLSR Routing Protocols in Static

Scenarios

Naga .V.Yedida, Rajesh Reddy Challa.

Center for Advanced Computer Sciences.
Email : {nxy4835,rxc2763}@louisiana.edu

Abstract: protocols specifically AODV, DSR (reactive) and OLSR-

Dynamic and reliable routing protocols are required in INRIA (proactive) protocols in a static environment with
the Ad-hoc multi hop wireless networks, as they have no RTS and CTS. Extensive study is going on Wireless Mesh
infrastructure (base station) and their network topology Networks, where mesh nodes form a static back-bone which
changes frequently. There are different protocols for are in ad-hoc mode. In such networks, it is important to
handling the routing problem in the Ad-hoc multi hop understand the performance of Ad Hoc routing protocols
wireless network environment. In this paper we focused on where every node is static. By this motive, we intend to
the two types of three popular routing algorithms Ad-Hoc compare few proposed Ad Hoc routing protocols and analyze
on Demand Distance Vector (AODV), Dynamic Source results in static environment. We need this work to study the
Routing (DSR) both being reactive routing protocols and behavior of each every protocol how they performing in
Optimized Link State Routing (OLSR), a proactive routing different scenarios and to find, which protocol performs
protocol. We evaluate and compare their performance better under a particular situation.
through simulation using QUALNET simulator. The The rest of the paper is organized as follows, in sec. II an
performance of the above mentioned protocols are overview of each protocol is given, in section III,
analyzed in static network environment with RTS and CTS experimental setup is described. In section IV, the results
in varying network size. obtained from the experiments are analyzed and in section V,
we draw conclusions.
I. INTRODUCTION

A mobile Ad-hoc network (MANET) is a collection of

II. BACKGROUND
nodes where the nodes will self configure and self organize
themselves forming a wireless medium with out any The existing protocols for ad hoc routing can be
requirement of stationary infrastructure like base station. In categorized as table-driven and on-demand scheme. Table
these networks each node will not only act as a host but -driven routing protocols (such as OLSR) attempt to
also acts as a router. Due to mobility of nodes, the topology maintain consistent, up-to-date routing information from
of the network is dynamic that is, it changes most of the each node to every other node in the network. In these
time . Some examples where the possible use of Ad-hoc protocols, every node in network should maintain a table to
networks are in military, in emergency situation like store route information. Nodes are required to periodically
hurricanes, earth quakes, conferences etc. One of the main propagate route update messages throughout the entire
issue in Ad-hoc networks is to develop a routing protocol network for the change of network topology. On the other
which must be capable of handling very large number of hand, on-demand routing protocols create route only when
nodes with limited bandwidth and power availability. Also the needs arise. When a source node needs a route to a
they should respond quickly to the hosts that broken or destination node, it starts a route discovery.
newly formed in various locations. Many protocols have For our performance comparison study, we pick up two
been proposed to solve these problems in the ad-hoc on-demand protocols (AODV, DSR) and one probably most
networks. In this paper we mainly discuss the performance well-known table-driven protocol, OLSR.
comparison of the two classes of three wireless multi-hop
routing protocols; reactive protocols like DSR and AODV, A. AODV : The Ad hoc On-Demand Distance Vector
and proactive protocol, OLSR-INRIA. We focus on (AODV) algorithm enables dynamic, self-starting, multi hop
comparing the performance of both proactive and reactive routing between participating mobile nodes wishing to

1
establish and maintain an ad hoc network. AODV allows very high rates of mobility.[1] [9]
mobile nodes to obtain routes quickly for new destinations,
C. OLSR: OLSR is a proactive routing protocol for mobile
and does not require nodes to maintain routes to
ad hoc networks. The protocol inherits the stability of a link
destinations that are not in active communication. AODV
state algorithm and has the advantage of having routes
allows mobile nodes to respond to link breakages and
immediately available when needed due to its proactive
changes in network topology in a timely manner. The
nature. OLSR is an optimization over the classical link state
operation of AODV is loop-free, and by avoiding the
protocol, tailored for mobile ad hoc networks.
Bellman-Ford "counting to infinity" problem offers quick
convergence when the ad hoc network topology changes OLSR minimizes the overhead from flooding of control
(typically, when a node moves in the network). When traffic by using only selected nodes, called MPRs, to
links break, AODV causes the affected set of nodes to be retransmit control messages. This technique significantly
notified so that they are able to invalidate the routes using reduces the number of retransmissions required to flood a
the lost link. message to all nodes in the network. Secondly, OLSR
requires only partial link state to be flooded in order to
One distinguishing feature of AODV is its use of a
provide shortest path routes. The minimal set of link state
destination sequence number for each route entry. The
information required is, that all nodes, selected as MPRs,
destination sequence number is created by the destination
must declare the links to their MPR selectors. Additional
to be included along with any route information it sends to
topological information, if present, may be utilized e.g., for
requesting nodes. Using destination sequence numbers
redundancy purposes.
ensures loop freedom and is simple to program. Given the
choice between two routes to a destination, a requesting OLSR may optimize the reactivity to topological changes
node is required to select the one with the greatest by reducing the maximum time interval for periodic control
sequence number.[1][8][9] message transmission. Furthermore, as OLSR continuously
maintains routes to all destinations in the network, the
protocol is beneficial for traffic patterns where a large subset
B. DSR: The Dynamic Source Routing protocol (DSR) is a of nodes are communicating with another large subset of
simple and efficient routing protocol designed specifically nodes, and where the [source, destination] pairs are changing
for use in multi-hop wireless ad hoc networks of mobile over time. The protocol is particularly suited for large and
nodes. DSR allows the network to be completely self- dense networks, as the optimization done using MPRs works
organizing and self-configuring, without the need for any well in this context. The larger and more dense a network,
existing network infrastructure or administration. the more optimization can be achieved as compared to the
DSR has been implemented by numerous groups, and classic link state algorithm[2].
deployed on several testbeds. Networks using the DSR OLSR is designed to work in a completely distributed
protocol have been connected to the Internet. DSR can manner and does not depend on any central entity. The
inter operate with Mobile IP, and nodes using Mobile IP protocol does not require reliable transmission of control
and DSR have seamlessly migrated between Wlans, messages: each node sends control messages periodically,
cellular data services, and DSR mobile ad hoc networks. and can therefore sustain a reasonable loss of some such
The protocol is composed of the two main mechanisms messages. Such losses occur frequently in radio networks
of "Route Discovery" and "Route Maintenance", which due to collisions or other transmission problems.
work together to allow nodes to discover and maintain Also, OLSR does not require sequenced delivery of
routes to arbitrary destinations in the ad hoc network. All messages. Each control message contains a sequence number
aspects of the protocol operate entirely on-demand, which is incremented for each message. Thus the recipient
allowing the routing packet overhead of DSR to scale of a control message can, if required, easily identify which
automatically to only that needed to react to changes in the information is more recent - even if messages have been re-
routes currently in use[3]. ordered while in transmission.[4][7]
The protocol allows multiple routes to any destination
and allows each sender to select and control the routes used
in routing its packets, for example for use in load balancing III. EXPERIMENTAL SETUP
or for increased robustness. Other advantages of the DSR
protocol include easily guaranteed loop-free routing, We used Qualnet 4.0 simulator. Qualnet provides a
support for use in networks containing unidirectional links, scalable simulation environment for multi-hop wireless ad
use of only "soft state" in routing, and very rapid recovery hoc networks, with various medium access control protocols
when routes in the network change. The DSR protocol is such as CSMA and IEEE 802.11.Four different topologies
designed mainly for mobile ad hoc networks of up to about (4x4, 8x8, 12x12, 16x16) are analyzed to compare three
two hundred nodes, and is designed to work well with even protocols (AODV, DSR and OLSR-INRIA). The experiment

2
is repeated three times each with different seeds and up to 11000000 bits/sec each.
different Source-Destination pairs for a particular protocol
and topology. Same set of Source-Destination pairs are
chosen for all protocols for any given topology, which IV. RESULTS
ensures performance of protocols for same scenario
precisely. (for example if AODV under 12x12 topology Simulations have been conducted with varying the node
and seed 2 has a pair, node 1 communicating to node 2, density and Source-Destination pairs . Following are the
OLSR has same pair under similar topology and seed, but metrics used for evaluation.
for any given protocol and topology, the Sourced- Metrics – In comparing the routing protocols four metrics
Destination pairs are mutually exclusive for different are chosen.
seeds). The complete experimental setup is given in
Table1. Packet Delivery Ratio (PDR): It’s the ratio between the
number of packets received at the application layer of the
destination node to the number of packets sent from the
Static Simulation Parameters application layer on the destination node.
Routing overhead: The total number of routing packets
Simulation Time 5 Minutes transmitted by sending or receiving node which involved in
the session during the simulation.
MAC protocol 802.11 with RTS/CTS
Throughput: It’s the average number of messages
PHY-MODEL PHY802.11b successfully delivered per unit time i.e. average number of
bits delivered per second.
Node Placement Grid End to End delay: It’s the time taken for a packet to be
transmitted from the source node to the destination node
Grid Unit 250 meters which includes all possible delays caused by buffering
during route discovery latency, queuing at the interface
Transmission Range 290 meters queue, retransmission delays at the MAC, propagation and
transfer times. In the session during the simulation.
Mobility None
Packet delivery ratio is important as it describes the loss rate
Propagation channel 2.4 GHz of the packets which will effects the through put of the
frequency network that it can support.. This metric characterizes both
the completeness and correctness of the routing Protocol. [5].
Propagation-Limit -111.0 Routing overhead is an important metric for comparing these
protocols, as it gives the scalability of a protocol, the degree
PHY-RX-MODEL PHY802.11b to which it will function in congested and low bandwidth
environments and also influences the amount of battery
Propagation Path Loss Free Space consumed to send these packets. The protocols which send
Model large number of routing packets can increase the probability
of collision, increasing in delay time and also finally results
PHY802.11-AUTO-RATE- YES in dropping of packets. [5].
FALL BACK
The rest of the metrics are also important in fact these
metrics determine how the above two metrics i.e PDR and
Routing Overhead influence them.
Table 1. Simulation Parameters.
A. Packet Delivery Ratio – In fig.1, the packet delivery
ratio of three protocols under various node density scenarios
Constant Bit Rate sources are used, which send data at is shown. We observed that the when the node density is 4x4
2048 bits/sec continuously. Each session is started 5secs the packet delivery ratio of all the three nodes is same, when
after experiment is started. For each simulation, 50% of the node density is increased from 4x4 to 16x16 the packet
nodes are participating in CBR session as a sender or delivery ratio is decreasing for each and every protocol.
receiver. (For example, in 8x8 grid, 16 nodes will send data Relatively for each node density AODV performed well and
packets to 16 receivers). With PHY802.11-AUTO-RATE- has the highest PDR up to 12x12 node density but, In 16 x 16
FALL BACK set YES, the PHY802.11-DATA-RATE and case DSR has the highest PDR. The decrease in packet
PHY802.11-AUTO-RATE-FOR-BROADCAST can reach delivery ratio for each and every protocol with increase in

3
 1.2 increasing linearly and in DSR it is hardly increasing, though
AODV being a reactive protocol the routing over head is
1 very high almost 5 to 7 times greater than that of DSR due to
Average Packet Delivery Ratio

following reasons that AODV periodically transmits Hello

0.8 messages, AODV is timer based scheme , AODV updates its
cache irrespective of the route being valid or stale when that
0.6 route is not used for a certain period of time . Where as DSR
AODV
is not timer based activity it will update its route information
DSR
0.4 in the cache on reception of RERR packet or on detection of
OLSR-INRIA
link breakage. AODV maintains only the next hop per
0.2 destination. DSR maintains multiple routes per destination .
DSR has the access to gather more information where as
0 AODV has limited access of information and finally the
4x4 8x8 12x12 16x16
overhead is high in AODV due to absence of promiscuous
listening[1]. As DSR adopts this feature it can save multiple
Number of Nodes routes in its cache, which results in low generation of
Figure 1. Packet delivery ratio for three protocols vs variable nodes. overhead . Promiscuous listening sometimes can cause the
pollution if there are stale routes between the nodes but, in
our case the probability of the presence of stale routes is very
node density is due to increase in number of hops which in low because of the environment being static where the link
turn increases the routing over head for route discovery cannot fail due to node mobility. We observed that the over
and latency which ultimately results in dropping of the pay head of OLSR is increasing linearly with node density. This
load packets. Up to 12x12 node density AODV out is expected because of its nature of being a proactive
performs both OLSR and DSR this is because of AODV protocol, hence large over head is expected as the route
being reactive protocol and mostly uses the optimal path discoveries are initiated periodically. This number can be
with moderate overhead but , in 16x16 node density there reduced by introducing jitter in the topology information
is a drastic change in the routing over head of AODV packets, in order to unsynchronize them so that the collision
which may be the reason for its degraded performance.(in of the periodic routing over head packets is reduced.[4]
fig 2.) In 16x16 node density DSR out performed AODV 100000
and OLSR. The main reason for this kind of performance is
Average number of Overhead Packets

due to comparatively low over head generation of DSR , 90000

aggressive caching of routes and promiscuous listening by 80000
the neighbor nodes, so that there wont be any need of new 70000
route request to be sent as the routes are stored in the cache
60000
and also the probability of the routes becoming stale is low
as the network we simulated is static. However in the case 50000 AODV
of network being dynamic these wont be the results, DSR’s 40000 DSR
PDR may fall drastically because the probability of the 30000 OLSR-INRIA
presence of stale routes in the cache is high and more over
DSR protocol induces these stale routes into neighboring 20000
nodes cache which results in polluting these nodes. OLSR 10000
never performed well up to 12x12 node density but when 0
we look at its characteristics with the increase in no. of 4x4 8x8 12x12 16x16
nodes the fall of PDR is some what linear when compared
to AODV (exponential), DSR (almost linear) and more Number of Nodes
over OLSR performed better than that of AODV when the Figure 2. routing overhead vs No.of nodes.
node density is 16x16 this is because of its proactive nature
(OLSR is independent of network density).[6] In order to have accurate comparison of protocols we
computed an alternative metric to routing over head called
“Normalized Routing load” [1] . Normalized routing load is
B. Routing Over head – we compared the Protocols based defined as the number of routing over head packets required
on routing overhead taken in no. of packets as from fig per one successful packet delivered at the destination. We
2.we observed that the routing over head for the AODV is observed from the fig.3 that the characteristics for the
increasing drastically when the node density is increased normalized routing load and routing overhead are pretty
where as the routing over head in OLSR is almost much same but we noticed that the relative difference

4
between normalized routing load for each protocol is increasing delay of the packets. The end to end delay of the
smaller when compared to difference between average OLSR protocol is slightly higher than that of AODV protocol
overhead (refer fig 2). We also observed that the in every node density . AODV out performs OLSR in end to
Normalized routing load for AODV is almost end delay because the network is stationary. In case of
exponentially raising where as in routing over head dynamic networks, with increase in node density, OLSR may
comparison its rise is almost linear (refer fig 2) . Average perform better due to well updated routes (search latency in
routing overhead may not actually provide a realistic AODV protocol is high which is the cause of degraded
comparison because the packet size of all control messages performance [6]). As our simulation environment is static,
is not equal ( for example RREQ is of fixed size of 24 AODV has the upper hand than OLSR. We observed an
bytes but RRER is not fixed and depends on number of link ambiguity in case of 12x12 node that there is a rise in delay.
failures). In the three seeds which we ran for OLSR, 12x12 topology,
only one of the seed was providing a delay of around 19 sec
800 and other two contributed around 8 secs and 6 secs, making
700 an average of around 11 secs. The average time in network
queue was found to be high for few sessions in this case. Few
Normalized routing load per received

600
more simulations are required for much accuracy in the case
500
AODV of OLSR.
packet

400 DSR
50
OLSR_INRIA
300
45

Average End-to-End Delay (secs)

200
40
100
35
0
4x4 8x8 12x12 16x16
30
No.of Node s 25 AODV
20 DSR
Figure 3. Normalized routing load vs No.of nodes.
15 OLSR-INRIA

10
C. End to End delay – We observed that the end-to-end
delay (fig.4), for each protocol, is almost the same in 4x4 5
topology and very low. When the node density is 0
increasing, the end to end delay for DSR protocol is 4x4 8x8 12x2 16x16
increasing exponentially and for OLSR its near to linear
Number of Nodes
rise and for AODV, it is hardly increasing i.e. the growth
rate of AODV delay is very low when compared to DSR Figure 4. End-to-End delay vs No.of nodes
and OLSR. As DSR is having low routing over head when
compared to OLSR and AODV, it is expected to have
minimum delay time, but the graphs show the results 18000
contrasting to our assumption. This is due to large size of
16000
the DSR over head packet when compared to AODV and
Average Throughput (bits/sec)

OLSR. Though the number of routing packets in DSR are 14000

low, the size of the packet is high. So when the routing
12000
over head is compared in terms of number of bytes, DSR
may have slightly larger or equal number of routing over 10000
head bytes. This large routing over head packet causes the 8000
AODV
delay and latency of the pay load packets in DSR when DSR
compared to AODV. The route discovery latency is added 6000 OLSR-INRIA
in the end to end delay for AODV and DSR which use hop- 4000
wise path length as a metric to choose the routing path
2000
from various alternative paths. AODV is better than DSR
in this regard as AODV replies to the first arrived RREQ 0
packet and discards other RREQs which arrive later from 4x4 8x8 12x12 16x16
other sources. This automatically favors the least congested
Number of Nodes
route instead of the shortest path[1]. Where as DSR replies
to all the RREQs that arrived and it will be difficult for the Figure 5. Throughput vs No.of nodes
protocol to select the least congested path which results in

5
D. Throughput: We observed that the Throughput (fig. 5) V.CONCLUSION
for all the protocols is very similar in low density networks
and as density increases, DSR outperforms AODV and AODV suits applications where End-to-End delays are very
OLSR. Though AODV has good throughput than OLSR, critical. Considering the Over all performance of AODV, it
its exponential drop suggests that for any further increase performs well in low and medium node density where as for
in network density OLSR performs better than AODV[7]. high node density both OLSR and DSR performs well. DSR
In low density topology, the number of hops is less, which is selected for the traffic which is highly dominated with file
implies the overhearing concept in DSR doesn’t have transfers where delivery ratio and throughput are the critical
much significance, but is favorable as density increases. factors with less importance to end-to-end delay . OLSR
DSR because of promiscuous listening and aggressive also performs equally well compared to DSR in high node
route caching policy always has an edge in high density density but it suits when the traffic is application oriented
networks. In OLSR, even in case of static topology, the ,like streaming traffic, video and voice, where end-to-end
synchronization of TC messages may happen in long term, delay is a critical factor. In Ad Hoc networks, scalability is
because of which, collision occurs and performance is one of the major problems. Scalability can be defined as
degraded. This in practice is avoided using jitter[4], but whether the network is able to provide an acceptable level of
still the problem persists due to many other factors like service to packets even in presence of large number of nodes.
congestions and drops, both in queues and in the medium If there is sufficiently enough bandwidth and resources,
and at the same time the over head in the data packet that OLSR protocol always provide best service irrespective of
is to be routed will be the same as in case of conventional network size. Even if bandwidth is limited, scalability can be
AODV protocol. OLSR is independent of traffic or achieved by limiting control updates to locations close to
network density so any further increase in density, OLSR changes. DSR also proves to be better as network size
should outperform. increases. The caching policy is critical in DSR protocol and
improvements are needed. The performance of the reactive
protocols may be improved by merging the properties of
E. Improvements in existing protocols
AODV and DSR as mentioned in section IV E.
REFERENCES
Promiscous listening s toring
[1] Samir R.Das, Charles E.Perkins, Elizabeth M.Royer ,
jus t s ingle hop information 6 Performance Comparison of Two On-demand Routing
Protocols for Ad Hoc Networks. IETF RFC, No. 3561, Jul.
S D 2003.
1 2 3 4 5
[2] T. Clausen, Ed., P. Jacquet, Ed., Optimized Link State
Routing Protocol (OLSR), IETF RFC, No. 3626, Oct. 2003.
RREP using s ource routing
AODV + DSR [3] D. Johnson, Y. Hu, D. Maltz, The Dynamic Source
Routing Protocol (DSR)for Mobile Ad Hoc Networks for
Figure 6. improvement of existing Protocol Ipv4, IETF RFC, No. 3561, Feb. 2007.
[4] P.Jacquet, P.Muhlethaler, T.Clausen, A.Laouiti,
Each protocol has its pros and cons. A hybrid protocol to A.Qayyum, L.Viennot, Optimized Link State Routing
combine properties of protocols can provide better Protocol fo Ad-Hoc Networks. Hipercom Project , INRIA
performance. Consider fig.6. Say, source node 1 has sent a Rocquencourt.
RREQ and destination node 5 has sent a RREP. Now if
node 6 wants to communicate with node 5 then it has to [5]Josh Broch,David A.Maltz,David B. Johnson, Yih-Chun
send RREQ to node 5 and then wait for RREP from latter. Hu and Jorjeta Jetcheva, A Performance Comparison of
(AODV has property of just storing the next best hop Multi-Hop Wireless Ad Hoc Network Routing Protocols.
neighbor information. DSR has the property of source [6] Aleksandr Huhtonen, Comparing AODV and OLSR
routing and listening in promiscuous mode to cache the Routing Protocols.
over heard path for future exploitation.) If both these
properties are merged, then the node 6 (neighboring node [7] Lars Christensen, Gitte Hansen, The Optimized Link State
to node 3) can over hear the source routed reply from 5 Protocol Performance Analysis through Scenario Based
through node 3 and can store the information of the route to Simulations., Master's thesis, Spring 2001.
5 so that it can have route information of node 5 before [8] Charles E.Perkins Elizabeth M.Royer, Ad-hoc On-
hand which can save over head if in case it wants send a Demand Distance Vector Routing.
data packet to node 5.
[9] Lecture slides of Alvin AuYoung, A comparison of Ad-
Hoc Routing Protocols. CSE 291-B April 24,2003.