International Journal of Computer Applications (0975 8887)

Volume 8 No.5, October 2010

Congestion Aware and Adaptive Dynamic Source Routing

Algorithm with Load-balancing in MANETs
A. Valarmathi
Assistant Professor
Department of Computer Applications
Anna University Tiruchirappalli, INDIA

ABSTRACT
The explosive growth in the use of real-time multimedia
applications in mobile devices requires quality of service (QoS)
support for effective communication. Network congestion is the
main reason for packet loss, longer delay and jitter in streaming
multimedia applications. Most of the present routing protocols are
not designed to adapt congestion. In this paper, the original DSR
protocol is modified to define the occurrence of congestion by
monitoring and reporting multiple resource utilization thresholds
as QoS attributes and use multi-path-routing and load-balancing
during the periods of congestion to improve QoS in MANETs for
CBR multimedia applications. NS-2 simulation was extensively
carried out to evaluate the performance of the modified DSR in
terms of throughput, jitter and end-to-end delay. The results were
compared with the original DSR in random node movements
under varying traffic loads of 10, 15 and 20 CBR traffic flows to
completely assess the applicability of the modified DSR.
Simulation results showed that the proposed protocol can
significantly improve the network performance for low and
medium traffic load in multimedia based mobile ad hoc networks
by extending the network lifetime and providing fewer packet
losses than the original DSR.
Keywords: Mobile ad hoc network; DSR; Congestion; Multi-path
routing; Traffic load

1. INTRODUCTION
The frequent and unpredictable topological changes in MANET
render the task of routing a challenging task and MANET
inherently depends on the routing scheme employed to provide
expected Quality of Service (QoS). Routing protocols for Ad hoc
networks can be broadly classified into two categories as
proactive and reactive. Proactive (or) table-driven routing
algorithms employ distance vector based or link state based
routing strategies. However, the main drawback of these
algorithms is that the need for frequent table updation consumes
significant amount of memory, bandwidth and battery power [1].
Example of such protocols is Optimized Link State Routing
(OLSR) [2] and Destination Sequenced Distance Vector routing
(DSDV) [3]. In reactive routing protocols, a node searches and
maintains a route to another node only when it requires
transmitting data to that particular node. This on-demand reactive
approach minimizes bandwidth utilization compared to proactive
routing protocols. Ad-hoc On-Demand Distance Vector (AODV)
[4], Dynamic Source Routing (DSR) [5], Temporally Ordered
Routing Algorithm (TORA) [6] and Associatively-Based Routing
(ABR) [7] are examples of such protocols. Our proposed protocol
is based on DSR. It operates on two mechanisms such as route
discovery and route maintenance. All these protocols rely on
single-path routing algorithms, where all traffic to a destination is

RM. Chandrasekaran
Professor
Department of Computer Science and Engineering
Annamalai University, Chidambaram, INDIA
required to be routed through a single successor and when a link
become congested, the entire traffic has to be rerouted. Routing
protocols for MANET can also categorized as congestion-unadaptive routing and congestion adaptive routing. Existing routing
protocols belong to the first group where congestion nonadaptiveness in routing may result in longer delay, higher
overhead and many packet losses for traffic intensive data like
multimedia applications. DSR does not make dynamic routing
adjustments based on memory, delay, number of service requests
or battery utilization information which often indicates certain
levels of usage and possible occurrences of congestion. Detection
of undesired paths enables the ability to implement route
admission control and to perform multi-path routing scheme
which could further improve the performance of the network in a
significant manner.
In the present paper, we propose a modified DSR protocol to
monitor congestion or multiple resource utilization levels, notify
the routing process of congestion and invoke multi-path routing
during the periods of congestion for streaming multimedia
applications. In addition, this paper also determine the effects of
various traffic loads(low, medium, high) models on the modified
routing protocol and compare the performance metrics of
throughput, jitter and end-to-end delay of original DSR protocol
with modified DSR protocol.

2. RELATED WORKS
The research on congestion for MANET is still in the early stage
and there is a need for new techniques. In this section, the
research work related to congestion control and multi-path routing
is presented. Round trip time measurements are used to distribute
load between paths in Multi-path Source Routing (MSR) [8]. A
distributed multi-path DSR protocol (MP-DSR) was developed to
improve QoS with respect to end-to-end reliability [9]. The
protocol forwards outgoing packets along multiple paths that are
subjected to an end-to-end reliability model. Split Multi-path
Routing (SMR) utilized multiple routes of maximally disjoint
paths which minimize route recovery process and control message
overhead [10-11]. The protocol uses a per-packet allocation
scheme to distribute data packets into multiple paths of active
sessions which prevents nodes from being congested in heavily
loaded traffic situations. Asis Nasipuri et al [12] showed how
intelligent use of multi-path technique in DSR protocol can reduce
the frequency of query floods. They also developed an analytic
modeling framework to determine the relative frequency of query
floods for various techniques. Lianfang Zhang et al [13]
established a network queuing model in MSR that could
incorporate cross-traffic among multiple paths. The load
balancing was considered as an optimization problem and the
results are verified with the existing model. Jinglun Shi et al [14]
presented an independent stability-based multi-path routing

International Journal of Computer Applications (0975 8887)

Volume 8 No.5, October 2010
scheme that is based on DSR. The stable multi-path embedded in
the routing policies is expected to enhance the performance by
decreasing rerouting overheads. Zafar Taha and Xian Liu [15]
demonstrated how the use of multi-path DSR can reduce the
frequency of route discovery process. The data traffic was
characterized through compound Log-normal distribution.
Stephen Mueller et al [16] specifically examined the issues of
multi-path routing in MANETs. They indicated that most of the
protocols proposed can only provide QoS in terms of specific
metrics, such as bandwidth, delay or reliability. However, it may
be necessary to develop mechanisms to support QoS in terms of
multiple metrics. Subramanian et al [17] proposed a Multi-path
Power Sensitive Routing Protocol (MPSR) based on DSR and
showed how an efficient heuristic based multi-path technique can
improve the mean time to node failure and maintain the variance
in the power of all the nodes as low as possible. A dynamic loadaware based load-balanced routing (DLBL) algorithm was
developed which considers intermediate node routing load as the
primary route selection metric [18]. This helps the protocol to
discover a route with less network congestion and bottlenecks.
When a link breaks because of the node mobility or power off,
DLBL provides efficient path maintenance to patch up broken
links to help to get a robust route from the source to the
destination. De Rango et al [19] proposed an energy aware multipath routing protocol by considering minimum drain rate as a
metric. An update mechanism and a simple data packet scheduling
among the energy efficient paths have also been implemented to
update the source route cache and for improving the traffic and
energy load balancing. Raghavandra et al [20] proposed
Congestion Adaptive Routing in Mobile Ad Hoc Networks (CRP).
In CRP every node appearing on a route warns its previous node
when prone to be congested. The previous node then uses a
bypass route bypassing the potential congestion to the first noncongested node on the route. Traffic will be split probabilistically
over these two routes, primary and bypass, thus effectively
lessening the chance of congestion occurrence.
The different multi-path schemes proposed earlier concentrates
few of the performance aspects. For example, the measurement of
jitter is very scarce, which is a critical performance metric for
multimedia traffic. In addition to that, a mere use of multi-path
routing schemes could adversely affect the network life as the
depletion rate of battery power will be higher. Hence, setting a
threshold value is critical to trigger multi-path routing as and
when congestion occurs to balance the network life and improve
the performance. Measuring only one parameter to ascertain
network conditions may present a false indication of network
performance. As a result, the measurement of multiple network
parameters may provide an enhanced indication of network
performance. Moreover, the effect of different traffic loads has
been evaluated to completely assess the applicability.

3. CONGESTION AWARE AND ADAPTIVE

DSR ALGORITHM WITH LOADBALANCING
The existing DSR protocol was modified to perform multi-path
routing with congestion aware and adaptive feature. In the
proposed protocol, the battery level and queue length are used as
the key resource utilization parameters.

3.1 Monitoring Congestion

Congestion should be defined for effective routing of packets in
the network. In the present paper, the congestion was assumed to
be in existence when queue length was near capacity or when
battery level fell below a pre-defined threshold. The proposed
protocol works similar to normal DSR protocol if the current
energy level is between 10 to 100 % of initial energy level and
queue length is less than 50. The modified DSR algorithm invokes
multi-path routing when it exceeds these threshold values.

3.2 Energy Calculation and Queue Length

The calculation of energy level is important to determine the
battery level of every node during active data transmission. In this
work, it was assumed that the battery level of a wireless node
decreased when the node initiated data transmission or when the
node forward packets. Our proposed algorithm calculates the
energy level of a node before forwarding data packet to the next
node. The queue length has been considered as another important
resource utilization parameter and the best indicator of congestion.
In a particular instance, when a data packet needs to wait in queue
for longer time, there is a possibility for unexpected delay in
transmission or dropping of packets.

3.3 Load Balancing

Whenever congestion occurs, the multi-path routing is invoked
which will select unequal cost paths stored in the secondary cache.
The modified DSR protocol transmits a RERR message when the
energy level reach 10 percent and the queue reach their capacity
of 50 packets. When the nodes in the downstream receive the
RERR message, it enables multi-path routing and searches their
primary and secondary caches for multiple routes to the intended
destination. During the period of congestion, the route admission
control ignores subsequent RREQ messages for 12 seconds, a 2
second more than the default flooding time of RREQ. This is
intentionally set for the routes to converge prior to the expiration
of this route admission control timer. After the expiration of the
route admission control timer, RREQ messages are processed
again and the protocol behaves as normal or as multi-path manner
depending upon the existence of congestion.

4. SIMULATION FRAMEWORK
4.1 Simulation Set-up
The Simulation of the network has been carried out using NS-2
version 2.30 [21-23]. The 802.11 MAC protocol is defined to be
the wireless channel. Each mobile node uses a Two-Ray Ground
radio propagation model with an Omni antenna. The initial battery
capacity of each node assumed to be 100%. The battery capacity
of a mobile node was decremented in a predefined manner by the
txPower and rxPower levels which remains constant throughout
the simulation. In order to conform to the IEEE 802.11b
specification, the dataRate_ variable was set to 11 Mbps and the
basicRate_ variable was set to 1 Mbps. These settings are
important to change the default NS-2 setting of 2 Mbps. The CTS
/ RTS process used in IEEE 802.11b within NS-2 normally
generate huge volumes of data that made processing the trace files
complicated. In order to restrict the processing of unnecessary
data, the RTSThreshold_ variable was set to 3,000 bytes which
restrict the processing of packets that have a size greater than
3,000 bytes. As per the definition, the simulation scenarios did not
generate packets larger than 512 bytes which is well below the set

International Journal of Computer Applications (0975 8887)

Volume 8 No.5, October 2010
RTSThreshold_variable. The random topology consists of 800 x
800 meter grid with 20 nodes. The packet sizes of CBR
multimedia streams were fixed at 512 bytes with each node had a
maximum queue size of 50 packets. The simulation time was set
to be 400 seconds and the speed of the node is 10 m/s. The
performance of the network was evaluated at different traffic
loads of 10, 15 and 20 CBR by keeping the pause time and data
rate constant at 4 seconds and 200 PPS, respectively.

multi-hop routing. This might hinder the possibility of selecting a

valid and effective multi-path for data transmission.

4.2 Performance Metrics

For streaming multimedia application, throughput alone cannot be
considered as performance indicator and that traffic load may
have a deceive effect on other parameters. Reide and Seide [24]
insisted on measuring packet loss, end-to-end delay and jitter to
further validate the network performance. Durkin [25] stated that
packet loss and delay are the most important measures for CBR
traffic. In the present paper, the performance metrics such as
average throughput, average jitter, average end-to-end delay and
percentage packet loss was calculated and evaluated for both
original DSR and modified DSR using the following relationships.

5.2 Average Jitter

4.2.1 Average Throughput

AverageThroughtput

Number of bytes received 8

kbps
Simulation time 1000

Figure 1. Average throughput of original and modifiedDSR at

different workload scenarios.

(1)

The average jitter of modified DSR is marginally higher in low

and high traffic load scenarios and is lower in medium traffic load
as shown in Figure 2. The increase in jitter values is primarily due
to the increase in traffic flow among the multi-path.

4.2.2 Average Jitter

Jitter is a measure of variation in delay across multiple packets
associated with a given traffic flow. In streaming multimedia
applications, only a small and limited amount of jitter is tolerable.
4.2.3 Average end-to-end delay
The average end-to-end delay is a measure of average time taken
to transmit each packet of data from the source to the destination.
Network congestion is indicated by higher end-to-end delays.

5. RESULTS AND DISCUSSION

5.1 Average Throughput
In low traffic load, the throughput of original DSR is 586 kbps as
shown in Figure 1. The throughput value should generally
increase as traffic load increases. However, in medium traffic load,
it increases to 639 kbps and then decreases to 496 kbps at high
traffic load. This is indicating the capability of the network in
transmitting data up to 15 CBR traffic flow without loss of much
data packets. An increasing traffic load beyond that limit leads to
loss of more data packets which ultimately result in reduced
throughput. The modified DSR shows inferior performance to
original DSR and exhibits a decreasing trend in throughput value
with increasing traffic load. In low traffic load, the throughput is
544 kbps, slightly lower than the original DSR and in medium
traffic load the throughput is 200 kbps lower than original DSR.
In high traffic load, the throughput is only 358 kbps. In a multipath environment, the traffic should get uniformly distributed to
achieve better throughput values whenever congestion exists.
Among the two congestion parameters, the chances of queue
length to exceed the threshold value are higher in this case.
However, the increased traffic flow might lead to congestion in
intermediate nodes which might make those nodes unavailable for

Figure 2. Average jitter of original and modified DSR at

different workload scenarios.

5.3 Average end-to-end delay

Figure 3 shows the average end-to-end delay of original and
modified DSR at low, medium and high traffic load scenarios.
The original DSR result in higher end-to-end delays ranging from
587 to 1014 ms. In original DSR, the need for transmitting data
over a single path irrespective of traffic flow resulted in drastic
increase in delay values. The use of multi-path in distributing the
traffic overcomes the problem which exhibit reduced delay values.
In different traffic load scenarios, the delay varies between 254
kbps to 619 kbps.

International Journal of Computer Applications (0975 8887)

Volume 8 No.5, October 2010
Conference on Electrical and Computer Engineering, 2000,1, pp.
479-483.
[9] Leung R, Jilei Liu, Poon E, Chan A-L C, Baochun Li. MPDSR: a QoS-aware multi-path dynamic source routing protocol
for wireless ad-hoc networks. In Proc. 2001 Local Computer
Networks, 26th Annual IEEE Conference, 2001, pp. 132-141.
[10] Sung- Ju Lee, Mario Gerla. Dynamic load-aware routing in
ad hoc networks. IEEE Conference on Communications, 2001, 10,
pp. 3206-3210.

Figure 3. Average end-to-end delay of original and modified

DSR at different workload scenarios.

6. CONCLUSION
In this paper, congestion aware and adaptive DSR Algorithm with
load-balancing for MANET is proposed. In random traffic load
scenarios, the throughput of modified DSR is marginally lower
compared to original DSR. This is indicating the least influence
of multi-path on traffic load. The jitter values also higher for
modified DSR, except for the medium traffic load where 4 ms
reduction in jitter was achieved. However, the modified DSR
exhibits excellent performance in reducing the end-to-end delay
by distributing traffic along multi-path. A huge reduction of 565
ms in delay was achieved under medium traffic load. The results
of the investigation imply that the modified DSR with multi-path
routing strategy is a viable alternative to original DSR for
streaming multimedia applications in MANET.

7. REFERENCES
[1] Macker J M, Corson, Park V. Mobile and wireless internet
services: Putting the Pieces together. IEEE Communications
Magazine, 2001, 36: 146-155.
[2] Clausen T, Jacquet P. Optimized link state routing protocol.
IETF RFC 3626, network Working Group, October 2003.
[3] Parkens C E, Bhagwat P. Highly dynamic DesinationSequenced Distance-Vector Routing (DSDV) for mobile
computers. Computer Communications Review, 1994, 24 (4):
234-244.
[4] Perkins C, Belding-Royer E, Das S. Ad hoc on-demand
Distance Vector Routing. July 2003, RFC 3561.
[5] Johnson D B, Maltz D A, Broch J. DSR- The dynamic source
routing protocol for multi hop wireless ad hoc networks. In Ad
hoc Network, Chapter 5, (Perkins C E eds.), Addison-Wesley,
2001.
[6] Corson S, Park V. Temporally-Ordered Routing Algorithm
(TORA) Version 1 Functional Specification. Mobile Ad Hoc
Network (MANET) Working Group, IETF, Oct. 1999.
[7] Toh C K. A novel distributed routing protocol to support adhoc mobile computing. In Proc. 15th IEEE Annual Int. Phoenix
Conference on Computers and Communication. 1996, pp. 480486.
[8] Lei Wang, Lianfang Zhang,Yantai Shu, Miao Dong. Multipath source routing in wireless ad hoc networks. 2000 Canadian

[11] Sung- Ju Lee, Mario Gerla. Split Multi-path Routing with

Maximally Disjoint Paths in Ad Hoc Networks. In Proc. Int. IEEE
Conf. on Comm., 2001, pp. 3201-3205.
[12] Asis Nasipuri, Robert Castaneda. Performance of Multipath
Routing for On-Demand Protocols in Mobile Ad Hoc Networks.
Mobile Networks and Applications, 2001, 6: 339-349.
[13] Linifang Zhang, Zenghua Zhao, Yantai Shu, Lei Wang, Yang
O W W. Load balancing of multipath source routing in ad hoc
networks. Int. Conference on Communications ICC 2002, 2002,
pp. 3197-3201.
[14] Jinglun Shi, Zhang Ling, Shoubin Dong, Zhou Jie. A
stability-based multipath routing algorithm for ad hoc networks.
In Proc. 14th IEEE Personal, Indoor and Mobile Radio
Communications, 2003,1, pp. 516-520.
[15] Zafar Taha, Xian Liu. A model for evaluating multi-route
routing in ad hoc networks. Wireless Telecommunications
Symposium, 2004, pp. 149-153.
[16] Mueller S, Tsang R P, Ghosal D. Multipath Routing in
Mobile Ad Hoc Networks: Issues and Challenges. Invited paper in
Lecture Notes in Computer Science, (Maria Carla Calzarossa,
Erol Gelenbe eds.), 2004.
[17] Anand Prabhu Subramanian, Anto A J, Janani Vasudevan,
Narayanasamy. Multipath Power Sensitive Routing Protocol for
Mobile Ad hoc Networks. wireless on-demand network systems.
LNCS 2928, 2004, pp. 171-183.
[18] Xiangquan Zheng, Wei Guo, Renting Liu, Yongchun Tian. A
New Dynamic Load-aware Based Load-balanced Routing for Ad
Hoc Networks. IEEE, 2004, 407-411.
[19] De Rango F, Lonetti P, Marano S. Energy-aware metrics
impact on Multi-path DSR in MANETs environment. Int. Symp.
on Performance Evaluation of Computer and Telecommunication
Systems, 2008, pp. 130-137.
[20] Duc A Tran, Harish Raghavendra. Congestion Adaptive
Routing in Mobile Ad Hoc Networks. IEEE transactions on
parallel and distributed systems, 2006, 17: 1294-1305.
[21] NS-2 with Wireless and Mobility Extensions, available via
web-site http://www.monarch .cs.cmu.edu.
[22] The ns Manual. The VINT Project. A Collaboration between
researchers at UC Berkeley, LBL, USC/ISI, and Xerox
PARC.December 13, 2003.
[23] The Network Simulator NS2. Availabl from:
http://www.isi.edu/nsnam/ns/, 2007.
[24] Reide N, Seide R. 802.11 (Wi-Fi) Networking Handbook.
first ed., CA, McGraw-Hill, 2003.
[25] Durkin. Voice-enabling the data network: H.323, MGCP, SIP,
QoS, SLAs, and security. first ed., Indianapolis, IN, Cisco Press,
2003.