Click here to load reader
Upload
azzedine-boukerche
View
218
Download
3
Embed Size (px)
Citation preview
WIRELESS COMMUNICATIONS AND MOBILE COMPUTINGWirel. Commun. Mob. Comput. 2004; 4:99–108 (DOI: 10.1002/wcm.167)
A performance evaluation of a pre-emptive on-demanddistance vector routing protocol for mobile ad hoc networks
Azzedine Boukerche*,y and Liqin Zhang
PARADISE Research Laboratory, SITE, University of Ottawa, Canada
Summary
Mobile ad hoc networks are useful for providing communication support where no fixed infrastructure exists or the
deployment of a fixed infrastructure is not economically profitable and movement of communicating parties is
allowed. Therefore, it is not possible to establish a priori and fixed paths for message delivery through the network.
Because of their importance, routing and packets dropped problems, mainly due to the path breaking, are among
the most studied problem in mobile and wireles ad hoc networks. Multi-path protocols can be useful for the
purpose of balancing congestion and decreasing the delay, by routing packets along different paths. However, they
may allow only source-based load balancing decisions. In this paper, we present a pre-emptive ad hoc on-demand
distance vector routing protocol for mobile and wireless ad hoc networks. We present the algorithm, discuss its
implementation and report on the performance results of simulation of several workload models on ns-2. Our
results indicate that a scheme based on scheduling a path-discovery routine before the current in-use link breaks is
feasible and that such a mechanism can increase the number of packets delivered and decrease the average delay
per packet. It also improves the throughput (packet delivered ratio) and balances the traffic between different
source–destination pairs. Copyright # 2004 John Wiley & Sons, Ltd.
KEY WORDS: AODV routing protocol; congestion; mobile ad hoc network; mobile and wireless communiction;
ns-2
1. Introduction
The rapidly expanding wireless LAN and mobile ad
hoc networks coupled with the hardware miniaturiza-
tion and recent proliferation of hand-held portable
(e.g. laptop, palmtop) computers and personal digital
assistants (PDAs) have been the driving forces behind
the field of mobile computing ad mobile ad hoc
networks. Mobile ad hoc networking field has also
the potential to dramatically change the society as
workers become untethered from their information
sources and communication media.
Ad hoc wireless networks are composed of mobile
stations (nodes) communicating through wireless
channels without any fixed backbone support
[11,21]. Mobile hosts (MH) which can receive a
message directly from another mobile, within the
transmission range are said to be neighbors of MH.
A message sent by MH may be received simulta-
neously by all of its neighbors. Messages directed to
mobile hosts that are not within the sender’s transmis-
sion range must be forwarded by neighbors, which
should act as routers. A route may consist of any
number of hops.
*Correspondence to: Azzedine Boukerche, PARADISE Research Laboratory, SITE, University of Ottawa, Canada.yE-mail: [email protected]
Contract/grant sponsor: CRC-Canada Research Program, Canada Foundation for Innovation, and Ontario DistinguishedResearcher Award.
Copyright # 2004 John Wiley & Sons, Ltd.
Ad hoc networks are useful for providing commu-
nication support where no fixed infrastructure exists
or the deployment of a fixed infrastructure is not
economically profitable, and movement of commu-
nicating parties is allowed. Applications of ad hoc
networks include military operations (communication
in a hostile environment), rescue operations (rapid
deployment of a communication network where infra-
structures do not exist or have been damaged) and
sporadic happenings coverage (intense utilization of a
communication network for a very limited time).
Due to node mobility, routes connecting two
nodes may change. Therefore, it is not possible to
establish a priori and fixed paths for message delivery
through the network. Because of its importance, rout-
ing is the most studied problem in ad hoc wireless
networks. To this end, a number of routing protocols
have been proposed for ad hoc wireless networks
[5,6,9,10,14,16,17,19,22], which are derived from
either distance-vector [12] or link-state [13] based
classical routing algorithms.
There exist two broad types of protocols, namely
proactive and reactive. Proactive protocols continu-
ously construct routes so that they are immediately
available when needed. They include the destination-
sequenced distance vector (DSDV) protocol [19,20],
and its Randomized DSDV (R-DSDV) [7], the wire-
less routing protocol (WRP) [17], the temporally
ordered routing algorithm (TORA) [19] and the light-
weight mobile routing (LMR) protocol [22]. On the
other hand, reactive protocols construct routes only
on demand, which include the ad hoc on demand
distance vector (AODV) protocol [17], and the dy-
namic source routing (DSR) protocol [10]. The zone
routing protocol (ZRP) is a hybrid of proactive and
reactive protocols [9].
Due to the frequent changes of topology in the ad
hoc network, minimizing the number of packets
dropped is an important issue. In this paper, we are
concerned with the congestion problem and packets
dropped that are mainly due to the path break in the ad
hoc network where the topology changes frequently,
and the packet delay due to finding a new path and the
transfer time. Multi-path protocols can be useful for
the purpose of balancing congestion and decreasing
the delay, by routing packets along different paths
[11,21]. However, they may allow only source-based
load balancing decisions. Therefore, there may exist
overloaded nodes even if sources equally distribute
their traffic over multiple paths to the same destina-
tion. This situation may be caused by several things
such as asymmetric topologies and asymmetric traffic
flows. Therefore, the use of some protocol may
become necessary in order for an overloaded node
along a path to influence the routing decision taken by
sources of the traffic through that node. The introduc-
tion of such a protocol may involve additional mes-
sage traffic to a node which is already congested, thus
making the scenario even more complicated.
In this paper, we present an extension to the AODV
routing protocol [2,17], which we refer to as PrAODV.
The PrAODV scheme consists of propagating the
route state information if the path is going to break
soon. It invokes a path rediscovery routine from the
source according to the received information about the
path which is contained in the route reply packet or
according to a warning message received at the
source. The purpose of our pre-emptive protocol is
to find a new path before the existing path breaks,
thereby switching to this new path. Our main focus
in this paper is to study the feasibility and the effec-
tiveness of the control mechanism in the proposed
protocol.
The remainder of the paper is organized as follows.
Section 2 presents the basic AODV routing protocol.
In Section 3, we present our PrAODV ad hoc routing
protocol, and outline the congestion control policy
based on the information about the path, followed by a
discussion of methods of how to get the route infor-
mation. In Section 4, we will report the simulation
experiments we have conducted to study the perfor-
mance evaluation of the pre-emptive AODV protocol
followed by a comparative study with the original
AODV protocol [2,17]. Section 5 concludes this
paper.
2. The Basic AODV Protocol
AODV [2,17] uses an on-demand-based protocol to
discover the desired path, while using hello packets to
keep track of the current neighbors. Since it is an on-
demand algorithm, it builds routes between nodes
only upon request by the source nodes. It maintains
these routes as long as they are needed by the sources.
It is loop-free and self-starting. Earlier studies have
also shown that AODV protocol scales quite well
when we increase the number of mobile nodes
[2,3,4]. AODV allows mobile nodes to obtain routes
quickly for new destinations and does not require
mobile nodes to maintain routes to destinations that
are not in active communication. It also allows mobile
nodes to respond to situations associated with broken
100 A. BOUKERCHE AND L. ZHANG
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
link and changes in network topology in a timely
manner, as we shall see later.
AODV uses a destination broadcast id number to
ensure loop freedom at all times, since the intermedi-
ate nodes only forward the first copy of the same
request packet. The sequence number is used in each
node, and the destination sequence number is created
by the destination for any route information it sends to
requesting nodes. It also ensures the freshness of each
established route. The route is updated if a new reply
is received with the greater destination sequence or
the same destination sequence number but the route
has fewer hops. Therefore, this protocol will select the
freshest and shortest path at any time.
� Path Discovery: When a route is needed, the source
invokes a path discovery routine. It has two steps:
the source sends out a request and the destination
return a reply. In what follows, we will discuss these
two steps.
� Step 1: Request Phase
When a route to a new destination is needed, the
node uses a broadcast RREQ to find a route to that
destination. A route can be determined when the
RREQ reaches either the destination itself or an
intermediate node with a fresh route to the destina-
tion. The fresh route is an unexpired route entry for
the destination associated.
The important step during the request phase is
that a reverse path from the destination to the source
can be set up. When a source node wants to find a
path to a destination, it broadcasts a route request
packet to its neighbors. The neighbors update their
information for the source node, set up backwards
pointers to the source node in their route tables, and
rebroadcast the request packet.
� Step 2: Reply Phase
When the request arrives at the destination or at an
intermediate node that has a path to that destination
a reply packet is returned to the source node along
the path recorded. While the reply packet propa-
gates back to the source, nodes set up forward
pointers to the destination, and set up the forward
path. Once the source node receives the reply
packet, it may begin to forward data packets to
the destination.
Figure 1 illustrates an example on how this scheme
works. In this example, a path from S to D is desired,
during the request phase, when the request arrives to
node a, it sets up a route entry whose destination is S,
[dest¼ s, next-hop¼ s]; and for b, it sets up for
[dest¼ s, next-hop¼ a]. Let say that b has a route
entry whose destination is D, and it is still active, the
request phase stops at b, and it returns a reply to S flow
the route that just set up, while returning the reply, it
sets up the forward route, for example, at node a, it
sets up the entry [dest¼ d, next-hop¼ b].
� Data Packet Forwarding: After the reply packet
returns from the destination, the source can begin
sending out the packet to the destination via the new
discovery path, or the node can forward any en-
queued packets to a destination if a reverse path is
set up and the destination of the reverse path is the
destination of the path.
A route is considered active as long as there are
data packets periodically traveling from the source
to the destination along that path. Once the source
stops sending data packets, the links will time out
and eventually be deleted from the intermediate
node routing tables.
� Dealing With the Broken Links: If a link breaks
while the route is active, the packets that are
flowing in that path can be dropped and an error
message is sent to the source or a local repair
routine will take over. If the node holding the
packets is close to the destination, this node invokes
a local repair route. It enqueues the packet and finds
a new path from this node to the destination.
Otherwise, the packets are dropped, and the node
upstream of the break propagates a route error
message to the source node to inform it of the
now unreachable destination(s). After receiving the
route error, if the source node still desires the route,
it can re-initiate a route discovery mechanism. How-
ever, we can add a pre-emptive protocol to AODV
and initiate a rediscovery routine before the current
routes goes down. In the next section, we will
discuss this pre-emptive approach in more detail.
Fig. 1. Path discovery.
MOBILE AD HOC NETWORKS 101
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
3. PrAODV: Adding Pre-emptive Protocolto AODV
In this section, we introduce the pre-emptive protocol
first, then we discuss how this pre-emptive protocol is
added into the original AODV, and finally we compare
the protocol with the original AODV protocol.
3.1. Pre-emptive Protocol
The pre-emptive protocol initiates a route rediscovery
before the existing path breaks. It overlaps the route
discovery routine and the use of the current active
path, thereby reducing the average delay per packet.
During the development of PrAODV development,
we have investigated several pre-emptive mecha-
nisms. In this paper, we will settle on the following
two approaches.
� Schedule a Rediscovery in Advance: In this ap-
proach, when a reply packet returns to the source
through an intermediate node, it collects the infor-
mation of the links. Therefore, when the packet
arrives at the source, the information about the
condition of all links will be known, including the
minimum value of the lifetime of the links. Hence,
we can schedule a rediscovery Trediscovery time
before the path breaks.
� Warn the Source Before the Path Breaks: Some
mechanisms are needed to take care of finding
which path is likely to break. We can monitor the
signal power of the arrived packets as follows:
when the signal power is below a threshold value,
we begin the ping–pong process between this node
and its immediate neighbor nodes. This node sends
to its neighbors in the upstream a hello packet
called ping, and the neighboring nodes will respond
with a hello packet called pong. Such ping–pong
messages should be monitored carefully. In our
approach, when bad packets were received (or we
timeout on ping packets) a warning message should
be sent back to the source during the monitoring
period. Upon receiving a warning message, a path
rediscovery routine is invoked.
3.2. PrAODV Protocol
In our pre-emptive AODV protocol, we combine the
above two pre-emptive mechanisms, and add them to
the original AODV. PrAODV protocol is built based
on the following assumptions. Future work will be
directed to eliminate such assumptions.
� Congestion is Caused Only Due to the Path Breaks:
Other factors such as the limited bandwidth avail-
able on wireless channels and the intensive use of
resources in an ad hoc network cannot affect our
protocol. We assume that the network has enough
bandwidth available, and each node has enough
capacity to hold packets.
� The Route Discovery Time can be Estimated: We
estimate the discovery time by monitoring the
request and reply packets, as the sum of the time
required to process and monitor these two packets.
� We can Detect When a Link Breaks: By monitoring
the transfer time of packets between the two neigh-
boring nodes.
Based on the above assumptions, we present our
PrAODV as the following:
� Schedule a Rediscovery: We have added an addi-
tional field, which we refer to as a lifetime, in the
reply packet to store the information of the links
along the new discovered path. Each link associates
a time-to-live, a parameter that is set up in the
request phase. The lifetime field contains the mini-
mum value of the time-to-live of the links along a
path. The reply packet also has an additional field, a
timestamp, which indicates the time that the request
packet was sent out.
Let us now see how we compute RTT, the esti-
mated time needed to discover a new path. When
the reply packet arrives at the source node, it com-
putes RTT as (arrival time� timestamp), then sche-
dule a rediscovery at time (lifetime�RTT). We use
RTT as our estimated value for the time needed to
discover a new path.
� Warning the Source: Recall that a pre-emptive
protocol should monitor the signal power of the
receiving packet. In our approach, we monitor the
transfer time for the hello packets. According to
Reference [1], Pn ¼ P0=rn, where P0 is a constant
for each link and n is also a constant, we can see
that the distance, which connects to the transfer
time of the packet via each link, has a one-to-one
relationship to the signal power. Thus, by monitor-
ing the transfer time of packets via each link, we
can identify when the link is going to break.
In our protocol, when a node (destination) re-
ceives a real packet other than AODV packets, it
sends out a pong packet to its neighbor (source), i.e.
where the packet comes from. Note that there is no
need to send this pong packet if we can include a
field in the real packet to store the current time the
102 A. BOUKERCHE AND L. ZHANG
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
packet is sent. However, in the original packet, it
does not have such a field, hence, we need an
additional pong packet to get such information. In
both ping and pong packets, we include a field
containing the current time that the packet is sent.
When the source monitoring the transfer time is
greater than some specific value c1, it begins the
ping–pong process. If more than two pong packets
arrived with the transfer time greater than a constant
c2, the node will warn the source of the real packet.
Upon receiving the warning message, the source
checks if the path is still active. This can easily be
done, since each route entry for a destination
associates a value, rt_last_send, which indicates if
the path is in use or not. If the path is still in use, then
a rediscovery routine is invoked.
Theoretically, these two cases can take care of
finding new paths before current paths break. The first
case is quite helpful if the data transmission rate is
slow. If only one data packet is sent out per time-
expired period, then even the current route is still
needed. Note that this entry may not have packets to
go through it during the next time period. This can
happen if the interval between two continuous packets
is greater that the timeout period of the current route.
The second case can be used to find a new path
whenever a warning message is generated. However,
those two cases are also based on some assumptions,
such as the span of a link which can be estimated by
the time out of ping packets and the values of those
constants are accurate. Once we break those assump-
tions, the test results will be unexpected.
3.2.1. Type of Packets in Pr AODV
The packet types in the AODV original protocol are:
route requests, route replies, triggered requests, trig-
gered replies and hello packets. In the PrAODV
protocol, we have added the following packets:
� Pre-emptive Packet: Used to inform the down-
stream nodes that a route rediscovery is needed,
and those nodes will update the information of this
route entry from RTF_UP to RTF_PRE, so that
those entries will be ready for the second request
packet.
� Ping Packet: Generated whenever a node receives a
hello packet which is longer than some value. Once
the ping packet is generated, it will send to the
source of the hello packet. We chose to monitor
periodically the ping–pong process.
� Pong Packets: Response packets sent to the source
of ping packet.
� Warning Packet: Generated during the ping–pong
period as a result of a ping packet timeout.
4. Simulation Experiments
In this section, we present the simulation model and
discuss the simulation experiments we have carried
out to evaluate the performance of the PrAODV
protocol. We have implemented the algorithm using
ns-2, version 2.1b8 [18]. The implementation of Pre-
emptive protocol based on the CMU Monarch exten-
sion. The distributed coordination function (DCF) of
IEEE 802.11 for wireless LANs is used as the MAC
layer. The traffic sources are continuous bit-rate
(CBR). The source–destination pairs are spread ran-
domly over the network. By changing the total amount
of traffic sources, we can get scenarios with different
traffic loads. The mobility models used in our experi-
ments to investigate the effect of the number of nodes
are 670� 670 m field with different number of nodes.
Each node starts its journey from a random location to
a random destination with a randomly chosen speed
uniformly distributed between 0 and 20 m/sec. Once
the destination is reached, another random destination
is targeted after a pause. Varying the pause time
changes the frequency of node movement. In each
test, the simulation time is 1000 s. The experimental
results were obtained by averaging several trial runs.
4.1. Performance Metrics
In order to evaluate the performance of the PrAODV
ad hoc routing protocol, we chose the following
metrics.
� Throughput: (i.e. packet delivered ratio) is a ratio of
the data packets delivered to the destination to those
generated by the CBR sources.
� Average Delay per Packet: This includes all possi-
ble delays caused by buffering during route dis-
covery latency, queuing at the interface queue,
retransmission delays at the MAC and propagation
and transfer time.
� Packet Arrived: This is the number of packets that
are successfully transferred to the destination.
4.2. Simulation Results
In this section, we report on the performance results of
simulation of several workload models of several
MOBILE AD HOC NETWORKS 103
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
scenarios we have used to evaluate the performance
the pre-emptive ad hoc routing protocol.
Before we proceed further, let us first define some
terms we will be using in this paper.
� Traffic Pair: The pair of the source and destination
of a traffic.
� Traffic Rate: The rate of the traffic generated
(packet/s)
� Quiet Time: The time interval between two contin-
uous packets generated at the same source.
In general, in both AODVand PrAODV, the number
of nodes does affect their performance. We have
observed that with the pre-emptive protocol, PrAODV
can significantly improve the performance of the
throughput, the average delay per packet, and the
packet delivered. The pause time and the packet rate
can also affect its performance. In the course of our
experiments, we have observed that PrAODV has
much better performance than AODV when the packet
rate is low.
Let us now turn to our results.
In our experiments, we set the traffic flow to ten-
pairs, the interval traffic rate to 4, and the pause time
to 600. As illustrated in Figure 2 our results indicate
that the number of nodes does not have any effect on
both protocols, AODV and PrAODV. However, the
pre-emptive AODV protocol indicates that it is more
stable than AODV, i.e. when we look closely at the the
standard deviation 9.036961 for AODVand 5.0943 for
PrAODV. This is mainly due to the fact that PrAODV
balances the traffic quite efficiently to a new path
before the current one breaks. The best performance is
observed when the number of nodes is set to 35, as
illustrated in Figure 2. In our experiments, we have
noticed that this type of behavior is always true with
only five traffic pairs as well.
In our next set of experiments, we wish to study the
effect of the number of traffic on the performance of
PrAODV. Based upon our initial experiments, see
above, we choose a scenarios with 35 nodes and five
traffic pairs. We also set the pause time at 600 s. Our
results are as indicated in Figure 3. In both protocols,
we observe an increase of the number of packet
arrivals as we increase amount of traffic. The through-
put obtained with PrAODV is lower when compared to
the AODV protocol, and the delay obtained with the
pre-emptive AODV protocol is lower when compared
to the original AODV. This is mainly due the pre-
emptive paradigm used in PrAODV and the number of
packets that are dropped during the simulation.
In Figure 4, we portray the results we have obtained
when we used 50 nodes, ten (source, destination)
pairs. Our results indicate that the pre-emptive proto-
col can balance quite well the traffic and the perfor-
mance for PrAODV is much more stable when
compared to the AODV protocol, see Table I.
In Figure 5, we portray the results we have obtained
when we use 35 nodes, and when the amount of traffic
Fig. 2. Effect of the number of nodes.
104 A. BOUKERCHE AND L. ZHANG
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
pair is set to five. As we can see, the performance of
the pre-emptive AODV protocol is much better than
AODV. Ths is mainly due to the the pre-emptive
mechanism with PrAODV that finds alternative paths
when links break.
From Figure 6, we portray the results we have
obtained when we vary the value of the interval/traffic
rate, and the interval between two contineous packet
from one source. As we can see from Figure 6, when
the quiet time is longer, the PrAODV has much better
performance than AODV. Since the longer the quiet
time, the more path broken can be rediscovery in
PrAODV protocol, and the performance can be much
better.
Fig. 3. The effect of the amount of traffic pairs.
Fig. 4. Ten traffic pair.
MOBILE AD HOC NETWORKS 105
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
In summary, the pre-emptive AODV invokes the
route rediscovery Tdiscovery before the current active
path breaks, so that we can find a new path before the
old one breaks and switch to it. Therefore, the packets
dropped can be decreased. Since the discovery routine
overlaps the communication process of the current
path, then the total delay will be decreased. After the
path switches to the new one, the packet will be
forwarded by using the new path, therefore, we can
Table I. The standard deviation for ten-pair traffic.
PTYPE Arrived Throughput Delay
AODV 13.768 34 0.910 771 1 0.036 882PrAODV 10.231 22 0.904 51 0.023 925
Fig. 5. The effect of pause time. Fig. 6. The effect of interval/traffic rate.
106 A. BOUKERCHE AND L. ZHANG
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
balance the load from the old path to the new path.
The load will be more balanced in pre-emptive AODV.
The overhead in the new protocol is almost the same
as the old one, except for the k ping–pong packets
monitoring, where k is a constant. Since the new
protocol is more balanced and the overhead is reason-
able, the pre-emptive AODV is easier to scale than
AODV.
5. Conclusion
In this paper, we have presented a pre-emptive on-
demand distance vector routing protocol for mobile
and wireless ad hoc networks. We have discussed the
algorithm and its implementation. We have also
reported a set of simulation experiments to evaluate
the performance of our scheme when compared to the
original AODV on-demand routing protocol.
Our results indicate that both AODV and PrAODV
protocols are easy to scale, and the number of mobile
nodes does not affect the performance of either of the
two protocols. Pre-emptive protocol objectives are to
find a new path before the current in-use path breaks
and switch to it. Therefore, the traffic can be balanced
to the new path. While our results indicate that this
protocol increases the number of arrival packets
within the same simulation time, it decreases the
delay per packet when compared to the original
AODV scheme.
In the future, we plan to extend our simulation
experiments and investigate further improvements of
our scheme and make it efficiently energy aware when
compared to previous protocols. We wish to enhance
our protocol by looking into the QoS parameters
based upon a multipath routing algorithm.
References
1. Preemptive routing in ad hoc network URL: http://opal.cs.binghamton.edu/nael/research/papers/mobicom01.pdf
2. AODV scalability study URL: http://moment.cs.ucsb.edu/AODV/AODVng_Presentations/lee.pdf
3. AODV protocol URL: http://www.cs.dartmouth.edu/snanda/adhoc/
4. Boukerche A. Simulation based comparative study of ad hocrouting protocols. 34th IEEE/ACM/SCS Annual SimulationSymposium, April 2001; pp. 85–92.
5. Boukerche A, Rogers S. GPS query optimization in mobile andwireless ad hoc networking. 6th IEEE Symposium on Compu-ters and Communications, July 2001; pp. 198–203.
6. Boukerche A, Das SK, Fabbri A. Analysis of randomizedcongestion control with DSDV routing in ad hoc wirelessnetworks. Journal of Parallel and Distributed Computing2001; 61: 967–995.
7. Boukerche A, Das SK. Congestion control performance of R-DSDV protocol for multihop wireless ad hoc networks. ACM/Kluwer Wireless Networks 2003; 9: 261–270.
8. Broch J, Maltz DA, Johnson DB, Hu Y-C, Jetcheva J. Aperformance comparison of multi-hop wireless ad hoc networkrouting protocols. Proceedings of ACMMOBICOM’98, Dallas,TX, USA, October 1998; pp. 85–97.
9. Haas ZJ, Pearlman MR. The performance of query controlschemes for the zone routing protocol. Proceedings of ACMSIGCOMM’98, Vancouver, BC, Canada, September 1998; pp.167–177.
10. Johnson DB, Maltz DA. Dynamic source routing in ad hocwireless networks. In Mobile Computing, Chapter 5, ImielinskiT, Korth HF (eds). Kluwer Academic Publishers: New York,1996; 153–181.
11. Lauer GS. Packet-radio routing. In Routing in CommunicationsNetworks, Chapter 11, Steenstrup ME (ed.). Prentice-Hall:New York, 1995; 351–396.
12. Malkin GS, Steenstrup ME. Distance-vector routing. In Rout-ing in Communications Networks, Chapter 3, Steenstrup ME(ed.). Prentice-Hall, 1995; 83–98.
13. Moy J. Link-state routing. In Routing in CommunicationsNetworks, Chapter 5, Steenstrup ME (ed.). Prentice-Hall,1995; 135–157.
14. Murthy S, Garcia-Luna-Aceves JJ. An efficient routing proto-col for wireless networks. Mobile Networks and Applications1996; 1: 183–197.
15. Network Simulator Version-2 URL: http://www.isi.edu/nsnam/ns/
16. Park VD, Scott Corson M. A highly adaptive distributed routingalgorithm for mobile wireless networks. Proceedings of IEEEINFOCOM’97, Kobe, Japan, April 1997; pp. 1405–1413.
17. Perkins CE. Ad hoc on demand distance vector (AODV) rout-ing. IEFT Internet Draft, available at: http://www.ieft.org/internet-drafts/draft-ietf-manet-aodv-02.txt, November 1998.
18. Pekins CE. Ad Hoc Networking. Addison Wesley Longman,Inc.: New York, 2001; 221–225.
19. Perkins CE, Bhagwat P. Highly dynamic destination-sequenceddistance vector routing (DSDV) for mobile computers. Pro-ceedings of ACM SIGCOMM’94, 1994; pp. 234–244.
20. Perkins CE, Bhagwat P. Routing over multi-hop wireless net-work of mobile computers. In Mobile Computing, Chapter 6,Imielinski T, Korth HF (eds). Kluwer Academic Publishers:New York, 1996; 183–205.
21. Ramanathan S, Steenstrup ME. A survey of routing techniquesfor mobile communications networks. Mobile Networks andApplications 1996; 1: 98–104.
22. Scott Corson M, Ephremides A. A distributed routing algo-rithm for mobile wireless networks. ACM/Kluwer WirelessNetworks 1995; 1: 61–81.
Fig. 6. Continued.
MOBILE AD HOC NETWORKS 107
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108
Authors’ Biographies
Azzedine Boukerche is a CanadaResearch Chair and an AssociateProfessor of Computer Sciences atthe School of Information Tech-nology and Engineering (SITE) atthe University of Ottawa, Canada.He is also the Founding and Direc-tor of PARADISE ResearchLaboratory (PARAllel, Distributedand Interactive Simulation ofLarge Scale Systems and Wireless
Mobile Networking). Prior to this he was an AssistantProfessor of Computer Sciences at the University of NorthTexas, and Director of the Parallel Simulations, Distributedand Mobile Systems Research Laboratory) at UNT. He alsoworked as a Senior Scientist at the Simulation SciencesDivision, Metron Corporation located in San Diego. He wasemployed as a Faculty at the School of Computer Science(McGill University, Canada) from 1993 to 1995. He spentthe 1991–1992 academic year at the JPL-California Instituteof Technology where he contributed to a project centeredabout the Specification and verification of the software usedto control interplanetary spacecraft operated by JPL/NASALaboratory. His current research interests include wirelessand mobile networks, wireless ad hoc and sensor networks,distributed and mobile computing, distributed systems,parallel simulation, large-scale distributed interactive simu-lation, and performance modeling. Dr. Boukerche has pub-lished several research papers in these areas. He was therecipient of the best research paper award at IEE/ACMPADS’97, and the recipient of the 3rd National Award forTelecommunication Software in 1999 for his work on adistributed security systems on mobile phone operations,and has been nominated for the best paper award at IEEE/ACM PADS’99 and at the ACM Modeling, Analysis andSimulation of mobile and wireless systems Conference2001. He served the General co-Chair of the principleSymposium on Modeling Analysis, and Simulation ofComputer and Telecommunication Systems (MASCOTS),in 1998, General Chair of the Fourth IEEE InternationalConf. on Distributed Interactive Simulation and Real TimeApplications (DS-RT2000), General Chair for the 3rd ACMConf. on Modeling Analysis, and Simulation of Wireless
and Mobile Systems (MSWiM’2000), as Program Co-Chairfor the Fifth IEEE International Conference on Mobile andWireless Computing and Communication (MWCN’03),ACM/IFIP Europar 2003, IEEE Wireless Local Networks(WLN’03), the 35th SCS/IEEE Annual Simulation Sympo-sium ANSS2002, and the 10th IEEE/ACM Symposium onModeling Analysis, and Simulation of Computer andTelecommunication Systems (MASCOTS’2002), the thirdInternational Conf. on Distributed Interactive Simulationand Real Time Applications (DS-RT’99), and ACMMSWiM’2000, as a Deputy Vice Chair of Wireless andMobility Access Track for ACM WWW 2002, as a GuestEditor for VLSI Design, the Journal of Parallel and Dis-tributed Computing (JPDC), ACM Wireless Networks(WINET), ACM Mobile Networks and Applications(MONET), and the International Journal of Wireless andMobile Computing. He was the main organizer of a SpecialSession on Performance Analysis of Mobile and WirelessCommunication Systems at the 7th IEEE HiPC Conference.He has been a member of the Program Committee of severalinternational conferences such as ICC, IPCCC, LNC, VTC,ICPP, MASCOTS, BioSP3, ICON, ICCI, MSWiM, PADSand WoWMoM, LWN, networking conferences. Dr. Bou-kerche serves as a Program Co-Chair for IEEE Workshop onWireless, Mobile Ad hoc and Sensor Networks (WMAN2004), and the IEEE Local Wireless Networks (LWCN2004). He also serves as an Associate Editor for the ACM/Kluwer Wireless Networks Journal, for the InternationalJournal of Parallel and Distributed Computing (JPDC), andthe SCS Modeling, Analysis and Simulation TRASNASC-TIONS, and on IEEE Task Force on Cluster Computing(TFCC) Executive Committee. He is a Steering Chair forIEEE DS-RT and ACM MSWiM Symposiums, and a mem-ber of the IEEE and ACM.
Linqin Zhang is a PhD Candidateat the SITE, University of Ottawa.She received her MSc Degree fromthe University of North Texas in2003. Her main research interestsare large-scale distributed systems,wireless and mobile ad hoc net-works.
108 A. BOUKERCHE AND L. ZHANG
Copyright # 2004 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2004; 4:99–108