An AODV Based QoS Routing Protocol for Delay Sensitive Applications in Mobile

Ad Hoc Networks

Rakesh Kumar, Anil K. Sarje, Manoj Misra

Department of Electronics and Computer Engineering
Indian Institute of Technology Roorkee -247667 India
Journal of Digital
{rkmlcdec, sarjefec, manojfec}@iitr.ernet.in
Information Management

Abstract: In order to provide quality delivery to delay sensi- routing protocols use minimum hop count as the main metric for
tive applications such as voice and video, it is extremely impor- path selection. In this paper, we propose an on demand delay-
tant that mobile Ad Hoc networks provide quality of service (QoS) oriented shortest path Quality of Service (QoS) routing protocol
support in terms of bandwidth and delay. Most existing routing (AODV-D) to ensure that delay does not exceed a maximum
protocols for mobile Ad Hoc networks (MANETs) are designed to value for each session between a pair of source and destination
search for the shortest path with minimum hop counts. However, mobile nodes. Quality-of-Service (QoS) is a desirable feature for
the shortest routes do not always provide the best performance, MANETs due to the growth of multimedia applications [23,30]
especially when there are congested nodes along these routes. and real time traffic. There are basic differences between the
In this paper, we propose an on demand delay based quality requirements for the transmission of bursty traffic and real time
of service routing protocol (AODV-D) to ensure that delay does traffic such as voice and video. Bursty traffic is error sensitive,
not exceed a maximum value. First in this paper, we compare while real-time traffic is delay-sensitive. Thus real time sessions
the performance of two prominent on demand routing protocols require bandwidth and delay guarantees. Traditional routing
AODV and DSR in stressful situation. As the performance of algorithms for Ad Hoc network can not meet this requirement.
AODV has been found better in our simulation than DSR, we Even though some approaches have been proposed to provide
decided to modify AODV routing protocol to provide quality-of- assurance in wired networks [21,24], wire-based QoS models
service. This protocol takes into consideration MAC layer chan- are not appropriate for Ad Hoc networks due to MANETs
nel contention information and number of packets in the interface inherent characteristics. In this paper, we propose an efficient
queue in addition to minimum hops. The protocol is implemented AODV based QoS routing protocol for providing end-to-end
and simulated using GlomoSim simulator. Performance com- delay guarantee in mobile Ad Hoc networks with IEEE 802.11
parisons of the proposed AODV-D protocol against AODV, and [17] as the MAC protocol. The solution consists of tracing routes
QoS-AODV is presented and shown that the proposed protocol in a reactive way by taking into account the QoS requirements
outperforms existing ones. in terms of delay associated with each flow. The performance
of the proposed algorithm is evaluated by simulation for different
Categories and Subject Descriptors mobility and traffic patterns.
C.2.2; [Protocol architecture]: C.2.4; [Distributed Systems
The remainder of the paper is organized as follows. Section
Distributed applications]
2 describes motivation behind development of the protocol.
General Terms: Mobile networks, Adhoc networks, Routing Related works are briefly presented in section 3. Brief descrip-
protocols tion of AODV and DSR protocols, simulation environment to
compare them under the scenarios we used, and results of
Keywords: Mobile Ad Hoc Networks (MANETs), AODV, QoS, simulation have been presented under section 4 and section 5
MAC, Performance Evaluation, GloMoSim respectively. Further, in section 6 we present the details of our
proposed AODV based delay sensitive QoS routing protocol
Received: (AODV-D) which include delay estimation, route discovery and
route maintenance. In Section 7, the simulation environment,
simulation results and analysis under various conditions of the
1. Introduction
proposed protocol are presented. Finally, section 8 concludes
In order to run delay sensitive applications such as voice and the paper and outlines future work.
video, it is extremely important that mobile Ad Hoc Networks [6]
provide Quality of Service (QoS) support in terms of bandwidth
2. Motivation
and delay. Most existing routing protocols for mobile Ad Hoc
Networks are designed to search for the shortest path with A key issue in MANET is that the routing protocols respond
minimum hop counts. However, the shortest route does not rapidly to topological changes in the network. Most of the
always provide the best performance, especially when there Mobile Ad Hoc networks use IEEE 802.11 [17] as the underlying
are congested nodes along the route. Delay aware routing MAC protocol. Current standard considers the shortest path
protocols make path selection between source and destination with minimum hop count for the route selection. Although this
based on the delay over the discovered links during route hop metric is easy to implement and is reliable in dynamic
discovery. Conventional routing protocols such as AODV, DSR environments, the queuing delay and the contention delay at
and OLSR [9, 10] are used to identify a path between the source the intermediate nodes are not taken into account for route
and destination nodes. But these protocols do not take into selection. Thus, a minimum hop path may sometimes incur a
consideration the node capabilities, queuing and contention higher end-to-end delay than some alternate paths. Moreover,
delays of intermediate nodes, which play an important role in routing protocols based on minimum number of hops can not
achieving QoS in modern day Ad Hoc networks. Moreover, such fairly distribute the routing load among mobile hosts [4,13].

Journal of Digital Information Management  Volume 8 Number 5  October 2010 303

 is very small and almost equal for each hop on the path. So,
queuing delay and MAC delay are considered as the two main
factors that accumulated the node’s delay.
Sheu and Chen [11] proposed a table driven delay-oriented
shortest path routing protocol (DOSPR). They analyzed the
relationship between the MAC delay and the neighbor number
in mobile ad hoc networks, and also provided an estimation
method for MAC delay. The average transfer delay of a packet
between a given source and destination includes MAC conten-
tion delay, buffer queuing delay and transmission delay. This
(a) Node X having one neighbor (b) Node Y having two neighbors protocol outperforms min-hop count routing. The total transfer
delay from source to destination for each packet has been
Figure 1. Effect of neighbor numbers on node delay reduced significantly. But as it uses table driven approach to
maintain access delay at each node, routing overhead will be
In IEEE 802.11 DCF, each node contends with its neighbor higher and also pose scalability problem.
nodes and also the neighbors of its neighbors in the medium
In [12,19], Sun and Hughes proposed an adaptive distributed
contention procedure. Consider Figure 1 to understand the
QoS routing scheme based on the prediction of the local per-
effect of contention delay and queuing delay [16]. In Figure
formance. They analyzed queuing delay by using two dimen-
1 (a), node X has only one neighbor whereas in Figure 1 (b),
sion finite-state Markov model. They gave the queuing delay
node Y has two neighbors. If Np is the number of packets at
distribution Pr (D<t).The average queuing delay is defined to
the mobile node p for transmission then, even though queuing
be the value D for which the delay distribution is larger than
delay of node X and node Y is identical, say four units each,
90%. Thus, the end-to-end delay of a path between two end
node Y experiences higher contention delay than node X as it
points can be estimated by adding all the node delays and link
has more number of neighbor nodes.
delays along the path. An H hops path has average queuing
Therefore packets passing through node Y experience higher delay given by [12]
node delay compared to node X as node Y has more number
H
of active neighbors (2 numbers) as compared to node X (1
number). But if we do not consider the contention delay, then
D(p)= ∑D
i =1
i (1)
packets passing through node X and Y will experience identical
node delay. Thus, MAC delay becomes much larger if the routing –
algorithm keeps routing other packets to pass through a heavily where Di = average queuing delay at node i. Due to distributed
loaded node. Since the range of possible medium contention of structure of the routing scheme, this protocol is scalable and
a mobile node is wide, medium contention times can affect the adaptive to its node’s mobility.
end-to-end delay considerably. MAC layer contention informa-
tion provides estimation of neighbor nodes activities whereas 4. Performance Comparison of AODV and DSR
queue length gives a measure of traffic load at the mobile node Protocols (Stressful / Less Stressful Situations)
itself. An unbalanced distribution of traffic may lead to higher
Here, first we briefly describe AODV [3] and DSR [4] protocols
packet dropping rate and faster battery power depletion on
functioning.
certain mobile nodes. If route selection criterion becomes least
path delay with minimum hop count instead of simple minimum
4.1 Dynamic Source Routing (DSR) Protocol
hop count, then it may find a route with least traffic load. If it can
DSR [4] uses source routing and thus the sender has complete
find alternate route(s) before congestion, then it can maintain
knowledge of hop-by-hop route to the destination. These routes
the required QoS constraints throughout the session.
are stored in a route cache. Every data packet carries the source
route in its packet header. When a mobile node in the Ad Hoc
3. Related Works
network tries to send a data packet to a destination for which
Perkins et al [13] have extended the basic Ad Hoc on Demand it does not already have the route in its route cache, it initiates
Distance Vector (AODV) routing protocol [3] to provide quality a route discovery process.
of service (QoS) support in Ad Hoc mobile wireless networks.
To provide QoS (bandwidth and delay guarantee), packet 4.2 Ad Hoc on Demand Distance Vector Routing (AODV)
formats (route discovery) and routing table structure were Protocol
modified in order to specify the service requirements which must The AODV [3] is a reactive routing protocol and shares features
be met by the nodes forwarding a RREQ or a RREP packet. of both DSDV [9] and DSR [4] algorithm. AODV shares Dynamic
Since NODE_TRAVERSAL_TIME at a mobile node is only the Source Routing (DSR’s) on-demand characteristics in that it
processing time for the packet, the major part of the delay at also discovers route as and when needed by initiating a route
a node is contributed by packet queuing and contention delay discovery process. It maintains one entry per destination in its
at the MAC layer. Hence a packet may experience much more routing tables unlike in DSR, which maintains multiple route
delay than this when the traffic load is high in the network. This entries for each destination in its route cache. In the routing
gave a motivation to develop a delay sensitive protocol that not table, there is an entry for each active route. Whenever a
only accounts packet processing time at a node but also MAC source node wants to send data to a destination node, it
contention and queuing delay. The end-to-end delay of a path is broadcasts RREQ packet. RREP packet is generated by either
the summation of the node delay at each node plus the link delay an intermediate node, which has a valid route to the destination
at each link (i,j) along the path. Node delay includes the protocol or by the destination node. When an active link breaks, the
processing time, the queuing delay at node i for link (i,j) and upstream node of broken link broadcasts a route error (RERR)
MAC contention delay at node i. Link delay is the propagation message to the source. The source node invalidates the listed
delay on link (i,j). In wireless network, the propagation delay routes and initiates a route discovery process.

304 Journal of Digital Information Management  Volume 8 Number 5  October 2010

5. Performance 5.3 Simulation Results
The effect of node mobility and offered load were studied as
In this section, we briefly describe the simulation environment
per above simulation parameters. Five experimental runs were
and various parameters chosen to simulate the routing protocols
repeated for each configuration, with a different starting random
AODV and DSR.
seed value for each run.
5.1 Simulation Environment and Scenarios
5.3.1 Scenario I (Stressful Situation)
To compare the performance of AODV and DSR routing
In the first set of simulation, the traffic was generated by 40
protocols, simulation experiments were performed using a
constant bit rate (CBR) sources spreading the traffic uniformly
parallel discrete event-driven simulator, GloMoSim (Global
among all nodes to 40 destinations. There were 100 mobile
Mobile Information System Simulator) [5]. Table 1 describes
nodes in the network. The packet rate was fixed at 4 packets/s
the detailed setup used for our simulation. One of the interests
and the size of each packet was 512 bytes. In this scenario,
of this paper is to test the ability of AODV and DSR routing
packet delivery rate of AODV is 30 % higher than DSR at high
protocols to react under stressful situation (under heavy
mobility whereas performance of DSR is about 14 % higher
load and high mobility). We experimented with 500 seconds
at low mobility than AODV (Figure 2). The delays of AODV
of simulated time over a rectangular field. We also added
are smaller than DSR by a factor of about 4 for high mobility,
50 seconds at the beginning of the simulation to stabilize
however at low mobility the DSR showed lower delay than
the mobility model. So the first data packet was initiated
AODV (Figure 3).
after 50 seconds only. Each simulation ran for 550 seconds
of simulated time. We ran our simulations with movement Routing overhead of DSR was found lower than AODV by a
patterns generated for six different pause times: 50, 70, 110, factor of 26 times at high mobility (Figure 4). We also examined
170, 350, and 550 seconds. the behavior of these two protocols with increase in number
of CBR sessions (transfer rate 6 packets/second for each
session) from 5 to 40, and observed the fraction of received
Routing Protocols AODV, DSR
packets at the destination. In both the protocols, the number of
MAC Layer IEEE 802.11 DCF received packets starts decreasing as we increase the number
of sessions. But AODV outperforms DSR when the number of
Transmission bandwidth of each link 2 Mbps
sessions is more than 20 (Figure 5).
Radio signal transmission range 250m
Packet size 512 bytes AODV : 40 sources
Packet Delivery Rate DSR: 40 sources
Terrain size 1500m × 300m 100

Number of mobile nodes 50 and 100 90

80
Node placement Uniform
70
Packet delivery rate(%)

Simulation time 550 sec (real simulation 60

time = 500sec) 50

Mobility model Random Waypoint [Speed 40

(0-20 m/sec), 30
Pause Time 20
(50,70,110,170,350,550 10
sec)] 0
50 70 110 170 350 550
Packet rate 4 and 6 packets/sec Pause time (sec)

Traffic type Constant Bit Rate (CBR)

Figure 2. Data packet delivery rate at varying mobility for 100 nodes
No. Of CBR source destination 20 and 40
pairs (flows)
AODV, 40 Sources
Table 1. Simulation Parameter Values End-to-End delay
DSR, 40 Sources
0.
5.2 Performance Metrics 3
In comparing both the protocols, the following performance 0.2
metrics were considered. 5

0.
• Packet delivery ratio: The ratio of the number of data packets 2
Average delay (secs)

successfully delivered to the destinations to those generated

0.1
by CBR sources. 5

• End-to-end delay: The average time from the beginning of a 0.

1
packet transmission at a source node until packet delivery
to a destination. This includes delays caused by buffering of 0.0
5
data packets during route discovery, queuing at the interface
queue, retransmission delays at the MAC, and propagation 0
5 7 11 17 35 55
and transfer times. 0 0 0 0
P a use time ( se c s)
0 0

• Routing overhead: The number of control packets generated

for routing by each routing protocol. Figure 3. Packet delay at varying mobility for 100 nodes

Journal of Digital Information Management  Volume 8 Number 5  October 2010 305

 AODV, 40 sources
Routing Overhead AODV, 20 sources
DSR, 40 sources End-to-End Delay
DSR, 20 sources
140000
0.12
120000
0.1
100000
No. of Control Packets

80000 0.08

Average delay (s)

60000
0.06
40000

0.04
20000

0 0.02
50 70 110 170 350 550
P a u s e T i me ( s e c s )
0
50 70 110 170 350 550
P ause T ime (seconds)
Figure 4. Routing overhead at varying mobility for 100 nodes
network
Figure 7. Packet delays with varying mobility for 50 nodes network
A ODV
Fraction of Received Packets DSR
1
Routing Overhead AODV, 20 sources
0.9 DSR, 20 sources
0.8 25000
Fraction of Received Packets

0.7

0.6 20000

0.5
No. of Control packets

15000
0.4

0.3
10000
0.2

0.1 5000
0
5 10 15 20 25
No .o f Sessio ns
30 35 40 0
50 70 110 170 350 550
P a use T ime ( se c o nd s)
Figure 5. Fraction of received packets for 100 nodes network

Figure 8. Routing overhead at varying mobility for 50 nodes network

AODV, 20 sources
Packet Delivery Rate
DSR, 20 sources
A OD V
90 Fraction of Received Packets D SR
80
1
70
Fraction of Received Packets

0.9
Packet Delivery Rate(%)

60 0.8
50 0.7
0.6
40
0.5
30
0.4
20 0.3
10 0.2
0.1
0
0
50 70 110 170 350 550
5 10 15 20 25 30 35 40
P a use T ime ( se c o nd s)
N o. of Sessi ons

Figure 6. Data packet delivery rate at varying mobility for 50 nodes

Figure 9. Fraction of received packets for 50 nodes network

5.3.2 Scenario II (Less Stressful Situation) mobility is high. The delay experienced by AODV is lower than
In the second set of simulation, there were 50 mobile nodes DSR at high mobility but at low mobility DSR outperforms with
in the network and 20 CBR sources were transmitting data to AODV (Figure 7).
20 destination nodes at the rate of 6 packets/second. From
DSR demonstrates significantly lower routing overhead by a fac-
Figure 6, it can be observed that packet delivery rate of both
tor of 5 times at high mobility (Figure 8) as compared to AODV.
AODV and DSR remain almost same with respect to mobility.
When we increase the number of CBR sessions (transfer rate 4
However, the performance of AODV is slightly higher when the
packets/second for each session) from 5 to 40, the number of

306 Journal of Digital Information Management  Volume 8 Number 5  October 2010

received packets starts decreasing as we increase the number Earlier papers [3,13,16] consider only minimum number of hops
of sessions. But AODV slightly outperforms to DSR when the as route selection metric. For route selection, it considers only
number of sessions becomes more than 20 (Figure 9). those routes which have total path delay less than or equal to
that specified in the route request. For calculating path delay, it
5.4 Observations takes into account MAC delay [11] at each mobile node along
Packet delivery ratio and packet delay of AODV and DSR the path. For route maintenance, each mobile node in the path
remains almost identical irrespective of the mobility in less piggybacks the delay information on data packets, so that
stressful situation. However, the performance of AODV in terms destination mobile node can initiate alternate route discovery in
of packet delivery is quite better as compared to DSR in the advance of congestion. In this section, we describe our proposed
stressful situation at high mobility whereas DSR performs better protocol, which includes calculation of forwarding_delay (i.e.
than AODV at low mobility. In stressful situation, at high mobility node delay) at each mobile node, initiation of route discovery
DSR packet delay is higher as compared to AODV whereas and route maintenance processes.
DSR out performs AODV as pause time becomes large. The
performance of DSR starts degrading due to cache staleness 6.1 Calculation of Node Delay
as well as high flow of traffic and large size of network at high Figure 10 [11] depicts the state transition diagram of a
mobility. The benefit of caching routes is completely lost in mobile node MNi, which tries to transmit packets in IEEE
case of DSR at high mobility and larger number of flows in big 802.11 standard [17]. IEEE 802.11 works in two modes:
networks. However, at low mobility, route information in route DCF (Distributed Coordination Function) and PCF (Point
cache always remains up to date. Due to aggressive use of Coordination Function). Data transmission in DCF [17] mode is
route caching and multiple routes per destination, DSR always depicted in Figure 11. The forwarding_delay at a mobile node
shows a lower routing load than AODV. The chances of finding MNi, which includes MAC contention and transmission delay
an alternate route from the route cache in DSR avoids frequent is calculated using equation (2) given in [11].
route discovery thereby reducing routing load significantly. In
AODV, routing load is mostly due to route request packets while Ddelay
i
= P idle
i
(DIFS) × (DIFS + avg_bt + DA(i) +
in case of DSR, it is mostly by route reply packets. (1-P idle
i
(DIFS)) × (SIFS+DB(i))+(L/R) (2)
DSR outperforms AODV significantly in terms of routing over- where
head in low mobility. However, its performance deteriorates
rapidly when the situation gets stressful. This is due to the P idle
i
(t) is the probability that mobile node MNi detects no other
aggressive usage of source routing cache. During a route mobile node transmitting data during time interval t and is
discovery process, the source can learn several routes to its given by
destination. This enables the source node to switch to some P idle
i
(t) = e -lt,l is the aggregate arrival rate (including neighbor
other cached routes in case of the current route breaks, which nodes) at mobile node MNi.
significantly reduced the possibility to restart a route discovery
process. However, in stressful situations, it is more likely that all DA(i) is expected delay encountered in the Attempt state and
cached routes are already invalid, thus introducing unnecessary is given by
delay and extra traffic.
DA(i) = P idle
i
(slot) × (RTS+2 × SIFS+CTS)+
AODV sustains better in stressful situation. For one route dis- (1-P idle
i
(slot)) × (RTS+2 × SIFS+DB(i)) (3)
covery process, destination node only returns one route reply
(RREP) packet. This requires that the source node should DB(i) is expected delay encountered during backoff state and
restart route discovery process to resume the transmission is given by
whenever current route fails. This is beneficial in high mobility DB(i)=[1/{P idle
i
(DIFS) × P idle
i
(slot)] ×
situations, where movement of nodes can quickly invalidate
[P idle
i
(DIFS) × P idle
i
(slot)}] × [P idle
i
(DIFS) ×
current route and cached route entries.
(DIFS+avg_bt+RTS+2 × SIFS+ P idle
i
(slot) ×
6. Proposed QoS Routing Protocol CTS)]+[(1-P idle
i
(DIFS)) × X] (4)

Some authors [1,2,7,8] also carried out performance evaluation avg_bt is a random backoff time interval before transmission
in different scenarios of AODV, DSR and other Ad Hoc routing and is given by
protocols through simulation using ns2. Our observations 4
regarding these two protocols were found close to them. AODV avg_bt ∑ (P idle
i
(slot) × (1 – P idle
i
(slot)**n ×
n=0
protocol outperforms DSR on performance metrics such as 2**(n–1) ×W) + (1– P idle
i
(slot))**5×2**4×W (5)
higher packet delivery ratio (PDR), lower end-to-end delay in
high mobility situation. Moreover, many QoS routing protocols L and R are packet length and data rate respectively,
[13, 31, 32, 33, 34, 35, 36] designed are extension of AODV W is contention window size,
protocol, we therefore preferred to extend AODV routing protocol
over DSR to propose a reactive QoS routing protocol for delay X=RTS+3×SIFS + CTS + L + ACK
sensitive applications in mobile Ad Hoc networks which we call
ACK is length of acknowledgement packet
AODV-D. The protocol modifies and extends QoS-AODV [13]
to discover a route with least traffic and maintain the required In wireless links, the propagation delays are very small and
QoS delay constraint throughout the communication process. almost equal for each hop along the path. So, here we assume
This algorithm selects routes with least traffic and follows that the propagation delay is negligible.
alternate route method for route maintenance. The protocol
The AODV-D protocol has two phases:
estimates node delay dynamically and destination nodes
(i) Route discovery phase
monitor the healthiness of the paths by piggybacking delay
(ii) Route maintenance phase
information and thus selecting better route before congestion.

Journal of Digital Information Management  Volume 8 Number 5  October 2010 307

 currently processing the RREQ. When a route is required
but destination information is not available in the routing
table, the source node floods the RREQ packet to discover
a route. Every RREQ packet carries the source and des-
tination addresses, a sequence number, hop count, delay
parameter (max_delay) and Accumulated Value extension
(acc_delay). Initially the value of acc_delay is zero. If the
source node wants to discover a QoS route, it records the
maximum acceptable delay in max_delay field otherwise it
sets it to –1. A node, which receives a RREQ packet mea-
sures forwarding_delay as per equation (2) and records this
value in its routing table. The forwarding_delay is MAC delay
Figure 10. State transition diagram of a mobile node (MNi ) in at a mobile node. This value is maintained at every node
proposed protocol along the path being discovered in their respective routing
table. If forwarding_delay > (max_delay – acc_delay), it
immediately drops the RREQ packet, otherwise RREQ is
broadcast to its neighbors after acc_delay field of RREQ
is updated by adding in it forwarding_delay recorded in
the route table. Also intermediate node records the current
value of acc_delay in RREQ to the acc_delay field of that
node route table. Route entries are created for every pair
of source and destination, i.e., for each session of com-
munication since each session may have different delay
Figure 11. Data transmission in IEEE 802.11 (DCF mode) using requirement. This process continues until the RREQ packet
RTS and CTS reaches the destination node. Since the destination node
receives a set of RREQ packets from different paths, it waits
6.2 Route Discovery Phase for a small timeout to allow all RREQ packets to discover
To provide end-to-end delay QoS, extensions are done in routes. After selecting an optimal route with the lowest value
RREQ, RREP messages and the routing table structure of of acc_delay, the destination node unicasts a RREP packet
AODV protocol as shown in Figure 12. along the reverse route towards the source. RREQs received
after generation of RREP are also buffered and used for
RREQ contains two extra fields: acc_delay and max_delay.
route maintenance phase. In AODV, RREP packet can be
The max_delay extension specify a maximum bound on the
created either by destination node or an intermediate node
acceptable time delay experienced on any acceptable path
with a fresh enough route to a destination [3]. Unfortunately,
from the source to the destination. Accumulated Value ex-
RREP packet can only be generated by destination node
tension field (acc_delay) enables the measurements to be
in AODV-D, because an intermediate node is not likely to
accumulated for end-to-end delay. It provides information
have current enough information about whether the remain-
about the cumulative value that has been experienced by
ing nodes along the path to the destination can also satisfy
nodes along the path from the originating node to the node
the requested QoS. When an intermediate node receives a
RREP packet, it updates the acc_delay of its route table entry
Destination Destination Next Active neighbors with acc_delay value contained in RREP packet. Certain
Hop count Timeout
addess sequence no. hop of the route fields (see Figure 12 a) are added to each route table entry
AODV
corresponding to each node requesting QoS. These fields
are added to notify endpoint nodes in cases where QoS
Destination Destination Hop Next
sequnce no. Timeout Active neighbors of forwarding_delay acc_delay parameter value are agreed upon, but the desired service
Address count hop the route

AODV-D
qualities can no longer be maintained.

(a) Routing table entry structure The node also updates the fields of RREP packet in the same
way as done for a RREQ packet as described above. Route
Source Destination Source Destination Hop count
discovery process of AODV-D is illustrated in Figure 13.
Broadcast_Id
address address Seq no. Seq no.

AODV 6.3 Route Maintenance Phase

AODV-D tries to maintain the QoS delay constraint throughout
Source Destination Source Destination
Broadcast_Id
Hop
Session_Id acc_delay max_delay
address address Seq no. Seq no. count

AODV-D
RREQ packet with RREQ packet with RREQ packet with
(b) RREQ message format Max_Delay=100 Max_Delay=100 Max_Delay=100
Acc_Delay=80
Acc_Delay=0 Acc_Delay=30

Source Destination Destination

Hop count Time out 1 2 3
address address Seq no.
Source Intermediate Node B Intermediate Node C Destination
AODV Node A Forwarding_Delay=30 Forwarding_Delay=50 Node D

Source Destination Destination 4

Hop count acc_delay Time out 6 5
address address Seq no.
AODV-D RREP packet with RREP packet with RREP packet with
Acc_Delay=80 Acc_Delay=50 Acc_Delay=0
(c) RREP message format

Figure 12. Message format Figure 13. Route Discovery Process (AODV-D)

308 Journal of Digital Information Management  Volume 8 Number 5  October 2010

the session by selecting alternate path(s). During data
transmission, each mobile node appends the delay information Step 4: The RREQ packet is further broadcasted by
to the data packets. Each packet header is time stamped intermediate node as per step 3 till destination node is
when the mobile node receives a packet. Let ai and bi denote reached.
the arrival and departure time of the ith packet respectively at Step 5: If destination node receives RREQ packet satisfying the
node p. After the ith packet’s successful transmission at a node QoS delay parameter
p, the estimated average total node delay q pi , which includes Then buffer the route.
contention, queuing and transmission delays at node p, is
computed using equation (6) [16]: Step 6: If buffer time expires
Then select a route with minimal delay and unicasts the
q pi = (1 – a) × q p + a × (bi–1 – ai–1) (6) RREP with acc_delay ← 0 in the backward direction;
i-1

where i > 1, 0 ≤ a ≤ 1, and ai–1 and bi–1 are arrival and departure Step 7: If destination receives RERR packet with a RREPFAIL
time stamps of previous packet i-1. flag due to route repair fail,
Then it selects a fresh route, next better route, from
Each intermediate node adds its node delay computed as per buffer and unicasts RREP to the source;
equation (3.10) to the piggybacked acc_delay field of each data
packet. Thus, destination node monitors the route capacity to Step 8: Intermediate nodes updates acc_delay field of RREP
serve the QoS requirements of the session. If total path delay packet as
(acc_delay) reaches the maximum limit the destination selects acc_delay←(acc_delay+ forwarding_delay);
next better route from the buffered active routes, those routes Step 9: If RREP packet reaches to source node within RREP_
which are not expired. If buffer does not contain any fresh routes WAIT_TIME, and
then it generates route error packet (RERR) in advance of con-
gestion. When a link breaks, then AODV-D tries to rebuild the If (acc_delay <max_delay)
broken link by doing local repair. The upstream node of a broken Then the source node buffers the route and start sending
link sends a local repair request to find the node following the data;
next node along the route to the destination. This request packet Else source node S restart route discovery with new
includes the session ID, with the time to live (TTL) value set to session Id;
3, which limits the broadcast area of the local repair request. To
allow this mechanism, the node following the next node along Step 10: If S receives a fresh RREP with same session Id
the route is also recorded in each routing table. If a local repair Then divert data transmission through new route;
mechanism fails, then a route error (RERR) packet is set to
notify the corresponding source nodes that the link is broken. Route Maintenance
Those mobile nodes, which receive RERR packet, invalidate Step 1: Destination node D monitors the healthiness of the on
the associated route entry. going route by examining piggybacked acc_delay field
value of each data packet received and comparing it with
6.4 Pseudocode of AODV-D route delay constraints i.e. max_delay;
The steps of the proposed algorithm (pseudocode) are as
under. Step 2: It pick up the next optimal route from buffered active
routes and sends RREP to source node to use alternate
Route Discovery route in the situation of congestion in advance;

Step 1: If Source node S has data packets to send and no route Step 3: If a node receives link break
is known to the targeted destination Then upstream node of a broken link perform local
Then initiate a RREQ with route repair by sending a local route repair request with
acc_delay ← 0; and limited broadcast (TTL = 3) to find the node following
max_delay ← d; // where, d is an upper bound to delay. the next node along the route to destination;

Step 2: Each intermediate node after receiving RREQ packet, Step 4: If local repair successful
calculates its own node delay as per equation (1) and Then update the route table;
records in its routing table forwarding_delay field. Then Else send RERR to source and invalidate the associated
compare its node delay value (i.e. current forwarding_ route entry.
delay value) with the value of (max_delay – acc_delay)
of RREQ packet. 7. Simulation Model and Performance Evaluation
Step 3: If (max_delay – acc_delay) > forwarding_delay To evaluate performance of our proposed AODV-D protocol,
we used the GlomoSim simulator [5,20,25] and compared
{
its performance with AODV and QoS-AODV protocols. The
Then update acc_delay field of RREQ as acc_delay ← simulations were conducted on an Intel Pentium IV processor
(acc_delay+ forwarding_delay); at 3.0 GHz, 512 MB of RAM running WINDOWS XP.
Record acc_delay of RREQ in acc_delay field of its
routing table; 7.1 Simulation Setup
Then broadcast the RREQ to its neighbor nodes; The performance of AODV-D was compared with the best effort
AODV [3] routing protocol and QoS-AODV [13] protocol using
}
GloMoSim. The GloMoSim library is a scalable simulation
Else environment for wireless network systems. It uses parallel
Drop the RREQ packet; discrete-event simulation provided by PARSEC [20,26], which
is a C based parallel simulation language. We examined the

Journal of Digital Information Management  Volume 8 Number 5  October 2010 309

effect of node mobility and traffic load on the three protocols. • End-to-End Delay: The average time from the beginning of
Some simulation parameters common for these protocols are a packet transmission at a source node until packet delivery
given in Table 2 [11]. to a destination. This includes delays caused by buffering of
data packets during route discovery, queuing at the interface
queue, retransmission delays at the MAC, and propagation
Parameter Value and transfer times.
Slot time 20msec
7.4 Results and Analysis
SIFS 10msec We present in this subsection the performance of the AODV,
DIFS SIFS+2 × slot time=50msec QoS-AODV and AODV-D protocols for the various metrics
presented above.
RTS 352msec
CTS 304msec 7.4.1 Effect of Node Mobility
A simulation model consisting of 50 mobile nodes with 20 active
ACK 304msec sessions, each with 8 packets/second arrival rate and variable
MSDU (MAC Layer Data Unit) 1500 bits pause time was considered for simulation. To study the effect of
mobility, pause time was varied from 0 to 900 seconds with steps
Bandwidth 2 Mbps of 100 seconds. This section describes the impact of mobility
on the performance of AODV-D, QoS-AODV and AODV.
Transmission range 250m
Carrier sensing range 550m 7.4.1.1 Average End-to-End Delay
With high mobility, AODV-D has lesser end-to-end delay
Arrival Rate 8 packets/sec
compared to QoS-AODV but slightly higher delay than best-effort
CWmin 32 slots AODV (with pause time < = 200 ms). This is because of high
interference and frequent link breaks. As mobility decreases (with
CWmax 1024 slots
pause time > 200 ms), average end-to-end delay of AODV-D
PHY scheme DSSS also decreases and it starts outperforming with both AODV and
QoS-AODV. After 700 ms pause time, average end-to-end delay
Noumber of mobile nodes 50 remains almost constant. This is because destination nodes go
Terrain size 1500m × 300m for alternate paths if the congestion occurs during transmission.
The average end-to-end delay with variable pause time of mobile
Simulation time 900 seconds nodes for these three protocols is depicted in Figure 14.
Transmission radius 100m
Mobility model Random Waypoint Average End-to-End delay for 20 AODV
QoS-AODV
sessions at 8 packets/sec AODV-D
Traffic CBR 1

0.9
Packet size 512 bytes
0.8
Table 2. Simulation Parameters
Average End-to-End Delay

0.7

7.2 Movement and Traffic Model 0.6

(Seconds)

The random waypoint model [15] is used to model mobility of 0.5

nodes. In this, mobile nodes speed is kept in between 0-20 m/
second. We varied the traffic load and the degree of mobility in 0.4
the simulation. We varied the traffic load by varying the number 0.3
of sessions (i.e. active sources) to be 5, 10, 15, 20, 25 and 30.
0.2
The pause time was kept at 400 second. We control the degree
of mobility by varying the pause times as 0, 100, 200, 300, 400, 0.1
500, 600, 700, 800 and 900 and the number of session was 20.
0
The number of mobile nodes in the simulation was 50. Other 0 100 200 300 400 500 600 700 800 900
parameters are given in Table 2. We used Constant Bit Rate Pause Time (Seconds)
(CBR) source as the data source for each mobile node. Each
source node transmitted packets at the rate of 8 packets/sec Figure 14. Average end-to-end delay vs. pause time for 20 sessions
with a packet size of 512 bytes. at 8 packets/sec traffic rate

7.3 Performance Metrics 7.4.1.2 Normalized Routing Overhead

In order to investigate the performance of these protocols, we Routing overhead of AODV-D is similar to QoS-AODV but higher
used following performance metrics: than AODV. This happens because AODV-D uses alternate
• Packet Delivery Ratio (PDR): It is the ratio between the route strategy in advance of congestion for route maintenance
packets received at the destination and the packets gener- and delay information is piggybacked on data packets. There is
ated by the sources. a general principle that QoS routing generally incurs a greater
• Normalized Routing Overhead: It corresponds to the ratio overhead than best-effort routing due to the extra information
between the total control packets sent and forwarded, and the being disseminated. The normalized routing overhead, with
successfully received data packets. Each time a control packet variable pause time of the mobile nodes, for these three
is retransmitted, it is considered as a new control packet. protocols is depicted in Figure 15.

310 Journal of Digital Information Management  Volume 8 Number 5  October 2010

 Normalized Routing Overhead for 20 AODV
QoS-AODV
nodes is increased resulting in frequent route changes. This
sessions at 8 packets/sec AODV-D causes the increase of average end-to-end delay at high traffic
5
4.5 rate. Even though AODV-D estimates node delay dynamically,
4 average end-to-end delay is almost similar to AODV and QoS-
Normalized Routing Overhead

3.5 ­AODV. The average end-to-end delay with variable number of

3
sessions for these three protocols is shown in Figure 17.
2.5
2
Average End-to-End delay at 400sec AODV
1.5 QoS-AODV
pause time and 8 packets/sec
AODV-D
1
2
0.5
1.8
0
0 100 200 300 400 500 600 700 800 900

Average End-to-End Delay (seconds)

Pause Time (Seconds) 1.6

1.4
Figure 15. Normalized routing overhead vs. pause time for 20 1.2
sessions at 8 packets/sec traffic rate
1

0.8
AODV
Packet Delivery Ratio for 20 sessions QoS-AODV 0.6
at 8 packets/sec AODV-D
100 0.4

95 0.2

90 0
5 10 15 20 25 30
Number of Sessions
Packet Delivery Ratio (%)

85

80
Figure 17. Average end-to-end delay vs. traffic load with pause
75
time = 400 second and 8 packets/sec traffic rate
70

65
7.4.2.2 Normalized Routing Overhead
Normalized routing overhead of AODV-D is slightly higher than
60
AODV as it uses local link repairs and congestion control to
55 preserve the QoS requirements. At high traffic, as the load on
0 100 200 300 400 500 600 700 800 900 the system is more, nodes become congested. To preserve QoS
Pause Time (Seconds)
delay requirement, destination nodes send new RREP packets
to source nodes changing to new routes. This causes the
Figure 16. Packet delivery ratio vs. pause time for 20 sessions at 8 increase of routing overhead. The normalized routing overhead
packets/sec traffic rate with variable number of sessions for these three protocols is
shown in Figure 18.
7.4.1.3 Packet Delivery Ratio
With high mobility, AODV-D has lesser packet delivery ratio 7.4.2.3 Packet Delivery Ratio
compared to AODV, but same as QoS-AODV. This is because Packet delivery of AODV-D is better than AODV and QoS-
AODV-D estimates node delay dynamically. With high mobility, AODV at high traffic (number of active sessions > =15). This
interference from neighboring nodes become high. As mobility shows the efficiency of the AODV-D. Average end-to-end delay
decreases (pause time >=200 seconds), AODV-D outperforms
both AODV and QoS-AODV, because of its better route AODV
Normalized routing Overhead at 400sec
maintenance policy. Up to 600 ms pause time, packet delivery pause time and 8 packets/sec
QoS-AODV
AODV-D
ratio is gradually increased, after 600msec pause time it 6
becomes almost constant. At low mobility, link breaks are few
and routes are balanced. The packet delivery ratio with variable 5
pause time of the mobile nodes for these three protocols is
Normalized Routing Overhead

shown in Figure 16. 4

7.4.2 Effect of Traffic Load 3

In our model, different traffic rates were simulated using different
number of sessions (sessions are varied from 5 to 30 in steps 2
of 5). A simulation model consisting of 50 mobile nodes using
random waypoint mobility with 400 sec pause time and with 1
8 packets/sec arrival rate is considered for simulation. This
section describes the impact of traffic load on the performance 0
5 10 15 20 25 30
of AODV-D, QoS-AODV and AODV.
Number of Sessions
7.4.2.1 Average End-to-End Delay
Average End-to-End delay of AODV-D is almost similar to AODV Figure 18. Normalized routing overhead vs. traffic load with pause
at low traffic. As traffic increases interference with neighbor time=400 second and 8 packets/sec traffic rate

Journal of Digital Information Management  Volume 8 Number 5  October 2010 311

 Packet Delivery Ratio at 400 sec AODV processing delay at each mobile node along the path. Hence
QoS-AODV it does not provide accurate estimation of end-to-end delay.
pause time and 8 packets/sec AODV-D
100 In this paper, only QoS metric considered is end-to-end delay
for a QoS flow, as finding a route subject to multiple metrics is
90 inherently difficult and in many cases is considered to be an
NP-complete problem [28]. It can be extended to bandwidth
Packet Delivery Ratio (%)

80 and other resource reservation schemes. In our protocol, we

used IEEE 802.11 DCF as a MAC layer protocol. With right set
70
of parameters, IEEE 802.11e EDCF [27] can be used at this
60 place as a MAC layer protocol.

50 References
40
[1] Broch, J., Maltz DA., DB, Johnson., Hu YC., J. Jetcheva.
30
(1998). A Performance Comparison of Multi-Hop Wireless
5 10 15 20 25 30 Network Routing Protocols. Proceedings of the Fourth Annual
Number of Sessions
ACM/IEEE International Conference on Mobile Computing and
Networking (MobiCom’98), Dallas, Texas, USA, p. 25-30.
Figure 19. Packet delivery ratio vs. traffic load with pause
time=400 second and 8 packets/sec traffic rate [2] Das, S. R., Perkins, C. E., Royer., Belding-Royer, E.M.
(2001). Performance Comparison of Two On- Demand Routing
and normalized routing overhead is better at high mobility in Protocols for Ad Hoc Networks. IEEE Personal Communications
case of AODV-D than QoS-AODV. The packet delivery ratio is Magazine, Special Issue on Mobile Ad Hoc Networks, Vol. 8,
significantly better at high traffic rate as compared to AODV No. 1, p. 16-29.
and QoS-AODV. AODV-D tries to maintain the routes with
[3] Perkins, C. E., Belding-Royer, E.M. (2003). Ad Hoc On-
the required low delays. This makes the routes less loaded
Demand Distance Vector (AODV) Routing, draft-perkins-manet-
and resulting lesser packet drops as congestion decreases.
aodvbis-00.txt, Internet Draft.
Therefore packet delivery ratio increases. With little expense
of routing overhead, AODV-D is able to deliver the packets [4] Johnson, D., Maltz,D., Jetcheva, J. (2002). The Dynamic
at good percentage. The packet delivery ratio with variable Source Routing Protocol for Mobile Ad Hoc Networks. Internet
number of sessions for these three protocols is shown in Draft, draft-ietf-manet-dsr-07.txt, Work in progress.
Figure 19. AODV-D is considerably performing well and it is [5] Zeng, X., Bagrodia, R., Gela, M. (1998). GloMoSim: A Library
able to meet the delay constraint efficiently up to 20 sessions for Parallel Simulation of Large Scale Wireless Networks,
and above 400 seconds pause time. The average end-to-end Proceedings of the 12th Workshop on Parallel and Distributed
delay exceeds 400 ms when the number of sessions increases Simulations, p. 154-161.
more than 20.
[6] Basagni, S., Conti, M., Giordano, S., Stojmenovic, Ivan
8. Conclusion and Future Scope (2004). Mobile Ad Hoc Networking, A John Wiley & Sons, Inc.,
Publication.
In this paper, first a performance comparison of two on
[7] Jorg, David Oliver. (2003). Performance Comparison of
demand routing protocols (AODV and DSR) was presented
MANET Routing Protocols in Different Network Sizes, Computer
as a function of traffic load and mobility. As a result of our
Networks & Distributed Systems.
findings through simulation, it can be said that DSR performs
poor in a stressful situations as compared with AODV at high [8] Johansson, P., Larsson, T., Hedman,N. (1999). Scenario-
mobility even though the routing overhead is higher in AODV based Performance Analysis of Routing Protocols for Mobile
as compared to DSR. DSR starts outperforming as mobility Ad Hoc Networks, Proceedings of 5 th Annual ACM/IEEE
decreases. As the performance of AODV routing protocol International Conference On Mobile Computing and Networking,
has been found better in general than DSR, we decided to p. 195-206.
modify AODV routing protocol to propose a novel on demand [9] Abolhasan, Mehran., Wysocki, Tadeusz., Dutkiewicz,
AODV based QoS routing protocol which we call AODV-D for Eryk (2000). Review of Routing Protocols for Mobile Ad Hoc
delay sensitive application in mobile Ad Hoc networks. For Networks, First International Conference on Networking,
performance comparison, we considered various numbers of France.
sessions with different packet rates and mobility model. Our
proposed protocol discover routes based on path delay in [10] Royer, E.M., Toh, C.K. (1999). A Review of Current Routing
addition to hop count instead of hop count only and also route Protocols for Ad Hoc Mobile Wireless Networks, IEEE Personal
maintenance is more efficient than existing standard. AODV-D Communications, Vol. 6, Issue 2, p. 46-55.
estimates node delay dynamically instead of taking a constant [11] Sheu, S.T., Chen. J.H. (2001). A Novel Delay-Oriented
value as in the existing QoS-AODV [13] standard. From these Shortest Path Routing Protocol for mobile Ad Hoc Networks.
simulation studies, it is evident that our protocol outperforms Proceedings of IEEE ICC, Vol. 6, p. 1930-1934.
AODV and QoS-AODV under average mobility and traffic load.
[12] Sun,H., Hughes. (2003). Adaptive QoS Routing Based
These characteristics make the protocol suitable for reliable
on Prediction of local Performance in Ad Hoc networks,
real time data and voice transmission applications in Ad Hoc
Proceedings of IEEE WNCN.
networks under IEEE 802.11 DCF as MAC layer protocol. Our
proposed protocol provides more accurate estimation of end-to- [13] Perkins, C.E., Belding-Royer, E M. (2003). Quality of
end delay. It has provision to avoid a congested path by keeping Service for Ad Hoc On Demand Distance Vector Routing,
track of Accumulated delay value extension field (acc_delay). draft-perkins-manet-aodvqos-02.txt, Mobile Ad Hoc Networking
Earlier papers [3,13] for delay estimation only considered Working Group Internet Draft.

312 Journal of Digital Information Management  Volume 8 Number 5  October 2010

[14] Macharl a, Pradeep., Kumar, Rakesh.,Sarje, Anil. K. [27] Dou, Chie., Chen, Yang-Jie Wang Jia Yuan. (2003),
Sarje Misra, Manoj. (2008). A QoS routing protocol for. delay- The Performance Study of Contention-Based Differentiation
sensitive applications in mobile ad hoc networks. The Second Mechanisms in IEEE 802.11e MAC Layer, The 14th IEEE 2003
International Workshop on Intelligent Networks:Adaptation International Symposium on Persona1,lndoor and Mobile Radio
communications and Reconfiguration (IAMCOM 2008) held Communication Proceedings.
in conjuction with COSWARE 2008 in Bangalore, India, p. [28] Wang, Z., Crowcroft, J. Quality-of-Service Routing for
720-727. Supporting Multimedia Applications, IEEE J. Select. Areas
[15] Hyytia, E., Koskinen, H. (2005). Random Waypoint Model Commun., Vol. 14, No. 7, p. 1228–1234.
in Wireless Networks. Networks and Algorithms: Complextiy in [29] Kumar, Rakesh., Sarje, Anil. K. Misra, Manoj. (2006),
Physics and Computer Science, Helsinki, p 16-19. Performance Comparisons Of AODV And DSR Protocols
[16] Song, J.H., Wong, V., Leung, V. C. M. (2003). Load- In Stressful Situation Using Glomosim, Proceedings
Aware On-Demand Routing (LAOR) Protocol for Mobile Ad of 3rd International Conference ObCom-2006: Mobile,
Hoc Networks. Proceedings of IEEE Vehicular Technology Ubiquitous & Pervasive Computing, Dec 18-19, VIT
Conference (VTC-Spring), Jeju, Korea. University,Vellore,India.
[17] LAN MAN Standards committee of IEEE Computer Society, [30] Kumar, Rakesh., Misra, Manoj Sarje, Anil K. (2007), A
Wireless LAN Medium Access Control (MAC) and Physical Review of Quality of Service (QoS) Route Provisioning in Mobile
Layer Specifications, ANSI/IEEE Standard 80211, 1999. Ad Hoc Networks, Journal of Digital Information Management
[18] Crawley, E., Nair, R.,. Rajagopalan, B., Sandick, H. (1998). (ISSN 0972-7272), Vol. 5, Issue 1, February 2007, p. 32-47,
A framework for QoS-based Routing in the Internet, Networking Available on line.
Group, RFC 2386. [31] Gerasimov, I. Simon, R. (2002)., A Bandwidth-Reservation
[19] Sun, H., Hughes, H.D. (2003). A Cross-Layer Analysis of Delay Mechanism for On Demand Ad Hoc Path Finding. IEEE/SCS
and Packet Loss for Support QoS in Ad hoc Networks, ICC. 35th Annual Simulation Symposium, San Diego, CA. 27-33.

[20] Bajaj, L., Takai, M., Ahuja, R., Bagrodia, R., Gerla,M. (1999). [32] Zhu, C., Corson, M. S. (2002)., QoS Routing for Mobile
GloMoSim: A Scalable Simulation Environment, technical Ad hoc Networks. INFOCOM 2002, Twenty-First Annual
Report 990027, UCLA Computer Science Department. Joint Conference of the IEEE Computer and Communication
Societies, Proceedings, 2, 958-967.
[21] Blake, S., Black, D., Carlson, M., Davies, E., Wang Z
Weiss,W. (1998), An Architecture for Differentiated Services, [33] Chen, L., Heinzelman, W. B. (2005)., QoS-Aware Routing
IETF RFC 2475, DEC. based on Bandwidth Estimation for Mobile Ad Hoc Networks,
IEEE Journal on Selected Areas in Communications, 23 (3).
[22] Internet Engineering Task force (IETF) Mobile Ad Hoc
Networks (MANET)Working Group Charter, http://www.ietf.org/ [34] Renesse, R. D., Glassman, M., Friderikos, V., Hamid, A.
html.charters/manet-charter.html. A. (2004)., QoS Enabled Routing in Mobile Ad Hoc Networks.
IEE 3G 2004, London, UK.
[23] Chlamtac, I., Conti M., Liu J.J.N. (2003). Mobile Ad Hoc
Networking: Imperatives and Challenges, AdHoc Networks, [35] Li, Xuefei, Cuthbert, L. (2005)., Multi-Path QoS Routing of
Vol. No. 1, p. 13-64. Supporting DiffServ in Mobile Ad Hoc Networks. Proceedings
of the Sixth International Conference on Software Engineering,
[24] Braden, R., Clark D., Shenker,S. (1994), Integrated Artificial Intelligence, Networking and Parallel/Distributed
Services in the Internet Architecture, IETF RFC 1633. Computing.
[25] Glomosim, http://pcl.cs.ucla.edu/projects/glomosim/ [36] Asokan, R., Natarajan, A.M. (2008)., Performance
obtaining_glomosim.html Evaluation of Energy and Delay Aware Quality of Service (QoS)
[26] Bragodia, R., Meyerr, R. (1999). PARSEC: Parellel Routing Protocol in Mobile Ad Hoc Networks. International
Simulation Environment for Complex Systems, UCLA Technical Journal of Business Data Communications and Networking,
Report. Vol. 4, Issue 2.

Journal of Digital Information Management  Volume 8 Number 5  October 2010 313

 Authors Biographies

Rakesh Kumar received his B.E. degree in Computer Engineering from M.M.M. Engineering College Gorakhpur
(U.P.), India in 1990 and his M.E. In Computer Engineering from S.G.S. Institute of Technology and Science,
Indore, India in 1994. Since January 2005, he has been a PhD student in the department of Electronics and
Computer Engineering at Indian Institute of Technology, Roorkee, India. He is a life member of CSI, ISTE and
also a Fellow of IETE. His main research interests lie in Mobile Ad Hoc Routing, Quality of Service Provisioning,
MANET-Internet Integration and Performance Evaluation.
Anil K. Sarje is Professor in the department of Electronics & Computer Engineering at Indian Institute of Technol-
ogy Roorkee, India. He received his B.E., M.E. and PhD degrees from Indian Institute of Science, Bangalore in
1970, 1972 and 1976 respectively. He served as Lecturer at Birla Institute of Technology & Science, Pilani, for a
short period before joining University of Roorkee (now Indian Institute of Technology Roorkee) in 1987. Dr. Sarje
has supervised a large number of M.Tech. dissertations and guided several Ph.D. theses. He has published a
large number of research papers in the International and National journals and conferences. He has also served
as referee for many reputed Journals like IEE Proceedings, IEEE Transaction on Reliability, Electronics Letters, etc. He
has been on a number of AICTE and DOEACC committees. He was a member of All India Board of Information Technology
during years 2000-2003. He is a senior member of the Institute of Electrical and Electronics Engineers (IEEE). His research
interests include Distributed Systems, Computer Networks, Real Time Systems and Network Security.
Manoj Misra is Professor in the department of Electronics & Computer Engineering at Indian Institute of Technol-
ogy Roorkee, India. He received his B.Tech. degree in 1983 from H.B.T.I., Kanpur and M.Tech. from University of
Roorkee in 1986. He did his Ph.D. from Newcastle upon Tyne, UK and joined Electronics & Computer Engineering
Department, University of Roorkee (now Indian Institute of Technology Roorkee) in August 1998 as Assistant
Professor. Before joining University of Roorkee, he worked in DCM, CMC Ltd., New Delhi, H.A.L. Kanpur and
H.B.T.I. Kanpur. He has completed an AICTE funded project “A CORBA framework for distributed mobile ap-
plications”, as a co-Investigator with Dr. R. C. Joshi. Dr. Misra has supervised a large number of M.Tech. Dissertations and
guided several Ph.D. Theses. He has published a large number of research papers in International and National journals and
conferences. He is a member of the Institute of Electrical and Electronics Engineers (IEEE). His research interests include
Distributed Computing and Performance Evaluation.

314 Journal of Digital Information Management  Volume 8 Number 5  October 2010