Proceedings of the International Conference on Computer and Communication Engineering 2008 May 13-15, 2008 Kuala Lumpur, Malaysia

Evaluation Study of QoS-Enabled AODV

Nur Idawati Md Enzai, Farhat Anwar, Omer Mahmoud

Electrical and Computer Engineering Department
International Islamic University Malaysia (IIUM)
P.O. Box 10, 50728 Kuala Lumpur, Malaysia.
Email: NurIdawati.MdEnzai@gmail.com, farhat@iiu.edu.my, and omer_one@yahoo.com

Abstract support of any fixed infrastructure. The ability to

operate without the support of any fixed infrastructure
Due to the dynamic nature of Mobile Ad-Hoc as well as relative ease of deployment make ad-hoc
Networks (MANETs), there are many issues which networks an attractive choice for various applications
need to be tackled. One of the areas for improvement including military, disaster recovery etc.
is Quality of Service (QoS) routing. Various In ensuring quality of service (QoS) provisioning, a
approaches and enhancements have been proposed network is expected to guarantee a set of measurable
including QoS-AODV. It is proposed to enhance one of pre-specified service attributes to the users in terms of
the existing routing protocols for MANET namely end-to-end performance, such as delay, bandwidth and
AODV (Ad-Hoc On-Demand Distance Vector). In jitter [3]. When it comes to QoS routing, the routing
QoS-AODV the original AODV is extended by adding protocols have to ensure that the QoS requirements are
necessary new fields including maximum delay met [2]. Unfortunately unlike wired networks, it is a
extension and minimum bandwidth extension. Our challenging task to ensure Quality of Service (QoS)
work highlights the combination of both metrics: delay provisioning including routing in ad-hoc networks due
and bandwidth respectively. The main aim of this to the mobile and dynamic nature of the nodes [7].
paper is to conduct evaluation study on QoS-AODV This paper is organized as follows: Section 2
along with our proposed approach. NS-2 Simulator is presents overview of QoS-AODV and its operation. In
used in performing the simulation study in order to test Section 3, we summarize related work in this field as
its effectiveness. Various scenarios with different well as the rationale behind this work. Section 4
mobility parameters are taken into account in the explains the simulation environment and the important
simulations. Average Latency, Packet Delivery Ratio performance parameters and metrics. The results are
(PDR) and Normalized Overhead Load are the presented and discussed in Section 5, and finally
selected evaluation metrics. Our experiments study the conclusion is drawn in Section 6.
effect of number of mobile nodes, maximum node
speed and traffic load as well as the impact of II. BRIEF OVERVIEW OF QOS-AODV
connectivity maintenance methods employed namely: One of major ad-hoc routing protocols is AODV
HELLO messages and MAC layer feedback. (Ad-Hoc On-Demand Distance Vector). Some Internet
Engineering Task Force (IETF) drafts have proposed
I. INTRODUCTION ways to extend AODV to provide QoS support
According to [9], Mobile Ad-Hoc Network [8][11][12][13]. Route discovery in AODV is on-
(MANET) is defined as an autonomous system of demand and follows a route request/route reply query
mobile nodes which are free to move about arbitrarily. cycle. When a source is in need of a route to a
One such example is wireless communications devices. destination, it broadcasts a Route Request (RREQ)
The mobile nodes may be located anywhere: airplanes, control in search of a route. Nodes having a current
ships, trucks, cars, perhaps even on people or very route to the indicated destination respond by unicasting
small devices. Another definition is stated in [1] a Route Reply (RREP) to the source node.
whereby ad-hoc wireless networks are described as the In order to provide QoS, extensions can be added to
category of wireless networks that utilize multi-hop these messages during the route discovery process.
radio relaying and are capable of operating without the Several extensions are needed in the routing table
structure and the RREQ and RREP messages as

978-1-4244-1692-9/08/$25.00 ©2008 IEEE 1254

reported in [2] [12]. The additional fields to each route IV. PERFORMANCE EVALUATION
table entry corresponding to each destination are:
maximum delay, minimum available bandwidth as well A. Simulation Environment
as list of sources requesting delay and bandwidth The parameters used in evaluating the performance
guarantees. Special messages which are called QoS of the selected protocols are given in Table 1.
LOST messages are forwarded to all sources
potentially affected by the change in QoS parameter. TABLE 1. SUMMARY OF SIMULATION ENVIRONMENT
These are the sources to which a RREP with QoS Number of nodes: 10, 20, 30, 40, 50
extension has been forwarded before [11].
Simulation Area Size: 700x700 square metre
MAC Layer: IEEE 802.11b
III. RELATED WORK
Transmission range: 250 metre
Based on our findings, there are several approaches Mobility Model: Random–waypoint
and proposals to cater QoS provisioning in the existing (0 pause time)
AODV. Perkins et al. themselves revised the QoS- Maximum node speed
AODV internet drafts by adding more field extensions (5 m/s, 10 m/s,
and new formats [8] [13]. IETF is not the only party 15 m/s, 20 m/s, 25 m/s)
involved in enhancing AODV to provide QoS support. Traffic Load (5 packets/s, 10 packets/s,
QoS-AODV proposals by Gerasimov and Simon only 15 packets/s, 20 packets/s
refer to original AODV internet draft. While de 25 packets/s)
Renesse, Ghassemian, Friderikos and Aghvami Routing: AODV
proposed a QoS-AODV enhancement by referring to Simulation duration: 1000 s
QoS-AODV IETF proposal [11]. Traffic type: CBR
Unlike Perkins et al. who identified the format and
No: of simulations: 8 runs
extensions to provide QoS support in AODV,
Gerasimov and Simon proposed an approach which
operates within a TDMA (Time Division Multiple B. Performance Metrics and Scenarios
Access) network [4][5][6]. On the other hand, de The metrics used in evaluating the performance of
Renesse et al. enhanced AODV by basing their work the selected protocols are as follows:
on QoS-AODV internet draft [11]. The main idea of - Packet Delivery Ratio
enabling AODV routing decision based on QoS - Normalized Overhead Load
objects is to add extensions to the route messages - Average Latency
during route discovery. This work is based on Mac In order to evaluate the performance of the selected
802.11b bandwidth measurement and reservation [7]. protocols as well as study their behaviors, different
However, de Renesse et al. only implemented the scenarios will be taken into consideration with respect
extension of minimum available bandwidth into to different mobility model parameters. As can be seen
AODV route messages for QoS provisioning purpose. from Table 1, the factors that have been considered in
The implementation of delay-based quality of service this study include:
routing has not been conducted. - number of nodes
In this paper, our work is based on de Renesse et - maximum node speed
al.’s implementation. Instead of only extending AODV - offered traffic load
messages with minimum available bandwidth, The process of varying the selected factors is
maximum delay field will also be extended in our essential in order to study their impact on QoS-AODV
implementation. For our simulations, we use NS-2 with regards to the evaluation metrics. Our
(Network Simulator 2) version 2.27 on Linux platform experiments study the effect of number of mobile
[10]. The main idea in QoS-AODV is to enable AODV nodes, maximum node speed and traffic load in order
routing decision based on QoS objects by adding to test its capability to perform in frequent changing
extensions to the route discovery messages [8]. topology as well as its scalability in congested
network. The structure of our NS-2 simulations
executions is shown in Figure 1.

1255
 Therefore, apart from the simulation which employs
NS-2 link detection by underlying MAC layer, we also
(C++) Results Traces simulate QoS-AODV approach which makes use of
HELLO messages as well as our proposed
Numerical Data enhancement. Table 2 summarizes the notations used
Simulation Scenario PERL/AWK along with brief descriptions of our studied protocols.
(OTcl)

Traffic Mobility
A. Effect of Number of Nodes
Pattern Pattern Studying the impact of varying number of nodes is
essential in order to test the applicability of the
protocols in large as well as small populations.
Figure 1. NS-2 simulations execution structure
1.00

0.98

V. RESULTS 0.96

0.94

0.92
TABLE 2. OVERVIEW OF THE SIMULATED PROTOCOLS
0.90

PDR
0.88

Notation QoS Parameter Connectivity 0.86

Maintenance 0.84
AODV-B-HELLO
Method 0.82 AODV-B-LL
0.80 AODV-BD
AODV-B- Minimum Link-Layer 0.78
LL Bandwidth Feedback 10 20 30 40 50
Num ber of Nodes
Extension
AODV-B- Minimum HELLO
HELLO Bandwidth Messages
Figure 2. Effect of number of nodes on PDR
Extension
AODV-BD Minimum Link-Layer From Figure 2, we could see that AODV-BD
Bandwidth Feedback achieves slightly higher value than AODV-B-LL while
Extension outperforms AODV-B-HELLO notably. This is due to
Maximum Delay additional control in allowing traffic through the delay
Extension guarantee along with bandwidth guarantee. Moreover,
along with increase of potential routes, there will also
Our implementation particularly refers to QoS- be more routes which could satisfy both bandwidth and
AODV proposal in [12] where the main idea is to delay guarantees alike.
provide extensions to AODV messages namely: With more nodes, there are more alternate routes
maximum delay and minimum bandwidth. In the QoS- that can be selected in order to reach the desired
AODV implementation in [7] (also referred in [12]), destinations which causing more packets to be
only minimum bandwidth field extension has been delivered, explaining the increase in ratio of packet
considered while issues involving delay guarantee has delivery.
not been addressed. Our highlighted contribution in Meanwhile, Figure 3 shows the effect of number of
this study is the implementation of maximum delay nodes on average end-to-end delay. Average latency
extension in addition to the existing minimum undergoes drastic drop until the simulation reaches the
bandwidth extension. In short, there are two QoS amount of 25 nodes. The reason is that with the
metrics which have been considered in this paper increase of potential paths which simultaneously leads
namely bandwidth and delay. to more connectivity; the packets are not required to
A method called sequential filtering is adopted for stay in the buffer for a longer period of time at the
our route computation purpose. The metrics are source.
classified as primary and secondary metrics, whereas As can be observed from both Figure 2 and Figure
the primary metrics are computed first and the 3, when the number of nodes becomes higher; PDR
elimination of the subsets are based on the secondary and average latency achieves almost constant values.
metric until the required route or path is found [14]. Irrespective of number of nodes, the optimum paths

1256
that have been found will be maintained. Accordingly, Figure 5. As mobility increases, more routes will
there will be no need of routes reestablishment thus not become invalid and new requests are required. While
affecting the average latency and ratio of delivered the requests are propagating the network in search for
packets. a new route, buffers will get full at the originating
node and packets are dropped. As can be noticed from
0 .0 6
Figure 5, AODV-B-LL receives more packets
compared to the other two protocols.
Average End-to-End Delay (seconds)

0 .0 5

0 .0 4 0.93 AODV-B-HELLO
0.92 AODV-B-LL
0.91 AODV-BD
0 .0 3 A O D V -B -H E L L O 0.90
A O D V -B -L L 0.89
A O D V -B D 0.88
0 .0 2 0.87
0.86
0.85
0.84

PDR
0 .0 1
0.83
0.82
10 20 30 40 50 0.81
N um ber of N odes 0.80
0.79
0.78
0.77
0.76
0.75
Figure 3. Effect of number of nodes on delay
5 10 15 20 25
Maximum Node Speed (m/s)

1 .1

1 .0 A O D V -B -H E L L O

0 .9
A O D V -B -L L
A O D V -B D
Figure 5. Effect of mobility on PDR
0 .8
Normalized Overhead

0 .7

0 .6

0 .5 0 .0 9 5
0 .4 0 .0 9 0
Average End-to-End Delay (seconds)

0 .0 8 5
0 .3
0 .0 8 0
0 .2
0 .0 7 5
0 .1
0 .0 7 0
0 .0 0 .0 6 5
10 20 30 40 50
0 .0 6 0
N um ber of N odes
0 .0 5 5
0 .0 5 0
0 .0 4 5 A O D V -B -H E L L O
Figure 4. Effect of number of nodes on overhead 0 .0 4 0
A O D V -B -L L
A O D V -B D
0 .0 3 5

Generally, as there are more potential routes in the 0 .0 3 0

0 .0 2 5

network, more control packets need to be delivered in 5 10 15 20 25

order to establish the paths, explaining the rise in M a x im u m N o d e S p e e d ( m /s )

overhead especially for AODV-B-HELLO. Figure 4

clearly demonstrates this. This is due to its mechanism
Figure 6. Effect of mobility on delay
which periodically broadcast HELLO messages for
each certain interval compared to the absence of this As indicated in Figure 6, while the network
method in AODV-B-LL and AODV-BD. The similar becomes more mobile, the latency increases. As there
performance achieved by AODV-BD and AODV-B- is higher probability of link failures and route change,
LL shows that, the addition of delay guarantee does there are also needs to forward QoS-LOST messages
not affect the amount of control packets transmitted. to the sources which are affected. QoS-LOST message
B. Effect of Mobility is generated by a node along the path which detects
As nodes mobility could cause frequent topology that QoS could not be maintained anymore. This
changes, the adaptability of the protocols with regards special message will be sent to all nodes potentially
to this factor needs to be studied. affected by the QoS maintenance failure. This is
Increasing mobility deteriorates QoS-AODV due to essential in order to reroute the traffic. As a result,
the increasing amount of route changes which explains large amount of QoS-LOST messages are transmitted.
the decrease of delivery ratio as can be observed from The time taken to transmit these messages and

1257
subsequent path reestablishment leads to higher end- the buffer for a longer time in order to have a chance to
to-end delay. get through. This long period of time caused the
increase in latency as can be seen from Figure 9.
AO D V -B -H E L L O
AO D V -B -L L
AO D V -B D 0 .9 5
0 .3 0

0 .9 0
0 .2 5
Normalized Overhead

0 .8 5
0 .2 0
0 .8 0

0 .1 5
0 .7 5

PDR
0 .1 0
0 .7 0

0 .0 5 0 .6 5 A O D V -B -H E L L O
A O D V -B -L L
0 .0 0 0 .6 0 A O D V -B D
5 10 15 20 25
M axim u m N o d e S p eed (m /s) 0 .5 5
5 10 15 20 25
T r a ffic L o a d (p a c k e t/s e c o n d )

Figure 7. Effect of mobility on overhead

Figure 8. Effect of traffic load on PDR
Figure 7 shows the effect of mobility on overhead.
Overall, normalized overhead increases with the
increase of maximum moving speed of the nodes. In
higher mobility networks, a node which is on the route 0 .2 0
A O D V -B -H E L L O
for transmitting traffic flow has higher possibility to Average End-to-End Delay (seconds)
0 .1 8 A O D V -B -L L
A O D V -B D
move out of the transmission range of the upstream or 0 .1 6

0 .1 4
down stream nodes. The upstream nodes are nodes that
0 .1 2
transmit the packets to the moving node under 0 .1 0
consideration and the downstream nodes are those that 0 .0 8
receive packets from the moving node under 0 .0 6
consideration. In order to alert source nodes that there 0 .0 4

is a loss of one of the intermediate nodes on the route 0 .0 2

and to find a new route, more and more route 0 .0 0

5 10 15 20 25
discovery and route maintenance packets are sent with T r a f f ic L o a d ( p a c k e t /s e c o n d )
the increase of the maximum moving speed of nodes.
C. Effect of Traffic Load
Figure 9. Effect of traffic load on delay
Investigating the effect of varying traffic load is
vital as the protocols must have the mechanism to
tackle the situation when the network becomes 0 .2 8
AO DV-B-HELLO
congested. 0 .2 6
0 .2 4
AO DV-B-LL
AO DV-BD
As can be observed from Figure 8, whether we use 0 .2 2
Normalized Overhead

link-layer feedback or HELLO messages, the PDR 0 .2 0

0 .1 8
decreases with the increase of sending rate. That is 0 .1 6

because the increasing sending rate of packets leads to 0 .1 4

0 .1 2
increase in network traffic as well. When the 0 .1 0

maximum throughput of nodes cannot satisfy the on- 0 .0 8

0 .0 6
going traffic, queues at nodes begin to be full; the 0 .0 4

packets in the end of queues of nodes will be dropped 0 .0 2

both at source nodes and at intermediate nodes. 5 10 15 20 25

Traffic Load (packet/second)
The effect of traffic load on latency is illustrated in
Figure 9. Though small drop is experienced for lower
Figure 10. Effect of traffic load on overhead
sending rate, latency involving both methods increase
as sending rate is increased to 10 packets per second. On the other hand, with regards to overhead, since
As sending rate becomes higher, the buffer would AODV-B-LL and AODV-BD employs link-layer
be filled up very quickly. The packets need to stay in feedback, the normalized overhead is constantly at low

1258
values. This is because, the AODV routing packets are possible contributing factors could be due to additional
only sent during the route searching and maintenance restriction in finding the suitable QoS route as well as
periods without exchanging HELLO messages. The the consideration of bandwidth alone has the capability
low normalized overhead for both AODV-BD and to reduce delay though indirectly.
AODV-B-LL is shown in Figure 10. However,
AODV-BD’s normalized overhead is slightly higher REFERENCES
than AODV-B-LL due to additional restriction which [1] C.S.R. Murthy and B.S. Manoj: Ad Hoc Wireless Networks:
requires more control packets while establishing the Architectures and Protocols, Prentice Hall, (2004)
suitable route. [2] I. Jawhar, and J. Wu: Quality of Service Routing in Mobile Ad
Hoc Networks, in M Cardei, I Cardei & DZ Du (eds), Resource
Management and Wireless Networking, Kluwer Academic
VI. CONCLUSIONS Publishers.
Our experiments study the effect of number of [3] P. Stüdi: Quality of Service for Mobile Ad Hoc Networks,
Diploma Thesis, (2003)
mobile nodes, maximum node speed and traffic load [4] I. Gerasimov, and R. Simon: Bandwidth-Reservation Mechanism
on QoS enabled AODV in order to test its capability to for On-demand Ad Hoc Path Finding, IEEE/SCS 35th Annual
perform in frequent changing topology as well as its Simulation Symposium, pp. 27-33 (2002)
scalability in congested network. Meanwhile, the [5] I. Gerasimov, and R. Simon: Communication Support for
Tightly-coupled Distributed Mobile Applications, International
selection of evaluation metrics helps determining the Journal of Simulation, Vol. 3, no. 2, pp. 23-40 (2002)
reliability and ability of this protocol to perform in [6] I. Gerasimov, and R. Simon: Performance Analysis for Ad Hoc
real-time application. QoS Routing Protocols, IEEE MASCOTS Mobility and
Though AODV is the basis of all our simulated Wireless Access Workshop (MobiWAC), pp. 87 – 94 (2002)
[7] R. de Renesse, M. Ghassemian, V. Friderikos and A. H.
protocols, the connectivity maintenance methods along Aghvami: Qos Enabled Routing in Mobile Ad Hoc Networks,
with admission control mechanism employed led to Fifth IEEE International Conference on 3G Mobile
performance distinctions between these three strategies Communication Technologies, pp. 678-682 (2004)
as demonstrated in the simulation results. [8] C.E. Perkins, E.M. Royer and S.R. Das: Quality of Service for
Ad hoc On-Demand Distance Vector Routing, IETF Internet
In conclusion, utilization of HELLO messages Draft: draft-perkins-manet-aodvqos-02.txt. (2003)
could result to performance degradation with respect to [9] S. Corson, and J. Macker: Mobile Ad hoc Networking (MANET):
the selected evaluation metrics due to the slower time Routing Protocol Performance Issues and Evaluation
it takes in detecting connectivity among the nodes. In Considerations, RFC 2501 (1999)
[10] Ns Notes and Documentation, http://www.isi.edu/nsnam/ns
contrast, link-layer feedback which makes use of [11] C.E. Perkins, E.M. Royer and S.R. Das: Ad hoc On-Demand
underlying MAC layer mechanism, any link break Distance Vector (AODV) Routing, IETF Internet Draft: draft-
could be detected much faster. ietf-manet-aodv-05.txt. (2000a)
The proposed enhancement that we made involves [12] C.E Perkins, E.M. Royer and S.R. Das: Quality of Service for Ad
hoc On-Demand Distance Vector Routing, IETF Internet Draft:
the addition of delay guarantee instead of bandwidth draft-ietf-manet-aodvqos-00.txt. (2000b)
consideration alone. From the results obtained, we [13] C.E Perkins and E.M. Royer: Quality of Service for Ad hoc On-
could see that the proposed enhancement of combining Demand Distance Vector Routing, IETF Internet Draft: draft-
delay and bandwidth metrics performs almost similar perkins-aodvqos-00.txt. (2001)
[14] G. Aggelou: Mobile Ad-Hoc Wireless Networks from Wireless
to the protocol which considers bandwidth only. The
LANs to 4G Networks, McGraw-Hill, Pages 151 – 226. (2005).

1259