Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks

Full text

European Journal of Scientific Research
ISSN 1450-216X Vol.49 No.1 (2011), pp.6-17
© EuroJournals Publishing, Inc. 2011
http://www.eurojournals.com/ejsr.htm
Adaptive Service Discovery Protocol for
Mobile Ad Hoc Networks
Cynthia Jayapal
Department of Information Technology, Kumaraguru College of Technology
Coimbatore – 641006, India
E-mail: cynthia_clement@rediffmail.com
Sumathi Vembu
Department of Electronics and Communication Engineering
Government College of Technology, Coimbatore – 641013, India
E-mail: sumi_gct2001@yahoo.co.in
Abstract
We present an adaptive service discovery protocol that enhances the performance of
service discovery. The existing service discovery procedures, uses either centralized,
distributed or hybrid architectures. These architectures use different methods of service
registration, advertisement, selection, discovery modes and state maintenance to improve
the service discovery performance, but they use the conventional methods for selecting a
core node that aids in all the service discovery phases. Our main focus is to use an adaptive
core node election mechanism that changes whenever the load increases and is also robust
against network failures. This enhances the performance of discovery due to the reduction
in frequent handoffs. We use a distributed directory based service discovery mechanism
that operates in a proactive mode with service advertisements to the core node and selects a
provider based both on distance and service capability of the provider. Our simulation
results show that our adaptive service discovery scheme performs better in terms of service
discovery success ratio, control message over head, discovery delay and the number of
hand offs, when compared to conventional schemes.
Keywords: Ad hoc networks, service discovery, adaptive leader election, service
advertisement, service selection
1. Introduction
Mobile ad hoc networks are networks with mobile nodes that are characterized with mobility, energy
limitations and uses multi hop communication. All nodes in an ad hoc network can send, receive and
forward data packets. A service is any entity or resource that is provided by one node and shared by
other nodes. Service discovery protocols enable resource sharing among mobile nodes dynamically.
Service discovery (SD) protocols are network protocols, which allow automatic detection of devices
and services offered by the devices on a computer network [1]. Services are located based on user
requests. A service request may be satisfied by several providers, or a service may have several
candidate providers. To enable service coordination, the service provision framework should be able to
track a group of service providers (SP) and their services. The challenges in service discovery are node
mobility, frequent disconnections, data management and security. With rapid proliferation of mobile

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 7
devices, service provisioning is becoming one of the fundamental operations of an ad hoc network. The
function of a SP is to advertise the available services, register its service with a registry node, update
the service attributes and deliver the service to the requestor. The function of a service requestor (SR)
is to query the SP information and forward a request either to the registry node or directly to the
provider and receive the service from the SP. There are different mechanisms to improve the
performance of the discovery as described in section 2.
In directory based service discovery architectures, a SP registers or advertises the service with
its descriptions. The details of the descriptions and the providers are maintained by the directory node.
A SR can send a request either to the directory node or to the provider. Service coordination is an
important task that is performed by the directory node in the discovery process. The directory nodes
are responsible for maintaining and sharing the service information and acts as an agent for the
requestor and the provider. It also aids in service delivery. The geographic position-based service
provisioning makes use of the position of the mobile node is vital for both service discovery and
delivery. In this case, the directory node also takes the responsibility of maintaining the location of the
providers and the requestors. Most of the existing discovery schemes use the conventional methods of
selecting a core node that aids in all the service discovery phases. These methods result in frequent
handoffs and affect the performance of service discovery.
We propose an adaptive service discovery protocol (ASDP) that uses an adaptive core node
election mechanism to elect a node that change whenever the load increases and is also robust to any
network failures. We elect a directory node based on an eligibility factor that considers the battery
power of the directory node, its distance to the center of the geographical zone and speed. This makes
the directory node more stable, reducing the handoffs. We use a distributed directory based SD
mechanism where the service advertisement is done proactively, state maintenance is done in a soft
state and service selection is done based both on the distance and service capability of a provider.
We simulate the adaptive service discovery protocol and compare its performance with two
other discovery procedures that use node ID and hash value to select a directory node. The
performance evaluation shows that the performance of ASDP is better than the other two schemes.
The rest of the paper is organized as follows. Section 2 provides a review of literature in service
provisioning domain. Section 3 describes the geographic decomposition for service provisioning.
Section 4 describes the role of a leader node and the adaptive leader election mechanism. Section 5
describes the adaptive service discovery protocol. In section 6, we evaluate the performance of the
adaptive service discovery protocol and Section 7 provides the concluding remarks for this paper.
2. Literature Review
There are various service discovery architectures proposed in the literature that use different discovery
modes, service selection strategies and state maintenance methods [1] – [5]. In the directory less
architecture, the mobile nodes do not distribute their service descriptions to the other nodes in the
network. A SR node sends its request message to all the neighbor nodes by broadcasting. If one or
more SP nodes can satisfy the request, a response is sent back to the SR by all the SPs. Here, the
broadcasting leads to high bandwidth and energy consumption and is not scalable. In directory based
architecture, the SPs register their services with the directory nodes and the service information is
provided to the requestors, through these directories only. The crucial issue in directory based
architecture is the election of the directory node. A directory can be either centralized or distributed. In
centralized directory architecture, a central directory stores the descriptions of all the services available
in the network. SPs advertise their services to the central directory using a unicast message. To access
a service, the SR first contacts the central directory to obtain the service description, which is then used
to interact with the SP. This helps in the communication between the SPs and the SRs, but it is hard to
scale and leads to bottlenecks as the centralized directory limits its scope to devices within local SD
domain. Moreover the central directory nodes energy depletes as its load increases. In distributed
directory architecture, instead of a single directory node, multiple nodes share the responsibility [6].

8 Cynthia Jayapal and Sumathi Vembu
These directory nodes constantly communicate with each other to spread and replicate service
information. The directory nodes integrate to form the backbone of the network region. The SPs
register their services with the directory which is geographically closer to its location [7]. Preserving
service consistency and controlling communication costs are some of the major challenges involved. In
hybrid directory architecture, the SPs register their services with service directories if they locate any
in their vicinity [5]. If not, they simply broadcast service advertisements to all the nodes in the
network. Here the response may come both from the SPs and service directories which leads to high
message overhead.
A SR can obtain service information using the reactive, proactive or hybrid service discovery
modes. In reactive mode [3], the SR sends a request directly to the coordinator or the SP, when needed.
The request may be sent as a unicast, multicast or broadcast packet. This increases the network traffic.
In proactive mode [8] [9], the SP advertises or registers its service with the coordinator or the directory
nodes. The advertisement can be sent to all coordinators or to those within the SP’s domain. Service
advertisement can also be multicast to a selective group of directory or backbone nodes. In hybrid
mode, both proactive and reactive communication takes place. The SPs advertise their services
periodically to the backbone nodes and the SRs query the backbone nodes for service information,
when required. If the service is not available the request and reply are made directly. There are two
service discovery strategies that can be employed in this mode – greedy strategy and conservative
strategy [10]. In the greedy strategy both service advertisement and service requests are broadcast to all
nodes, while in conservative strategy only a set of nodes, generally the backbone nodes receive the
service advertisements and service requests.
There are various service selection methodologies to select a SP. A service query may result in
multiple SPs. A proper selection mechanism is needed to handle multiple responses. The technique
may either be manual or automatic based on some criteria like current load of a SP, bandwidth
available for the communication channel between the SP and the client, velocity of the SP. Moreover,
the service selection algorithm can be run either directly by the SR or can be distributed among the
backbone nodes. Service selection may either be route specific [11] or service specific [4]. In route
specific service selection, the SP which is reachable with a minimum number of hop counts from the
SR is chosen. Though the cost of communication is reduced, quality of service cannot be guaranteed.
In service specific selection, among the multiple SPs, the SP with the highest capacity to serve is
chosen. The effectiveness of service is given priority over route and cost of communication.
Service state maintenance is required to maintain service information in MANETs, owing to
frequent changes in service availability. Failure to do this will lead to inconsistency. Service state
preservation can be done using two methods, hard state maintenance and soft state maintenance. In
hard state maintenance, the SP must inform the coordinator node if it moves outside the region. This
de-registration of services is not always guaranteed due to frequent node movements and
disconnections. In soft state maintenance, every service is associated with a time limit, until which the
mobile node offers the service. For service state maintenance, polling or notifications can be used. In
polling, the SRs approach the SP to obtain service information. Polling can be performed either on
demand or proactively. In notification, the SP sends notifications about service state changes to
registered clients. Notifications lead to increase in the number of control messages and hence traffic
[12].
Leader election is used in number of applications such as key distribution, routing coordination,
sensor coordination, group communication, cluster coordination. The role of a leader is to control a
group of nodes periodically and to act as an agent on behalf of them. In wired network the leader can
be a static server with enormous processing capability and memory capacity. But in an ad hoc or sensor
network the role of a leader and the responsibility of a node to act as a leader may vary time to time.
The leader election problem is a problem of electing a unique leader periodically based on the nodes
current capability [13]. In ad hoc networks, nodes with similar characteristics are often grouped to form
a cluster that changes adaptively. For each cluster, a cluster head is elected periodically to act on behalf
of the cluster nodes. The change in leader is essential for an ad hoc node, due to the limited battery

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 9
power of the node. In sensor networks, the role of the leader or cluster head is even more important as
it performs data aggregation, which greatly reduces the traffic [14].
3. Geographic Decomposition for Service Provisioning
Scalable Geographic Service Provision Framework for MANETs (SGSP) [8] and Distributed Hashing
for Scalable Multicast in Wireless Ad hoc Networks (HRPM) [15] are the two service discovery
techniques; we use to compare our adaptive service discovery scheme. SGSP uses node ID, HRPM
uses hash value and our adaptive algorithm uses an eligibility factor to select a core node within a
geographic zone. SGSP provides a self-configuring, distributed, hierarchical structure. It uses scalable
membership management to manage the dynamic collection of service nodes and their services. SGSP
is based on geographic unicast routing, in which every node is aware of its position and the
intermediate nodes make packet forwarding decisions based on local topology only. The entire network
is divided into manageable square zones virtually. Every zone has a local coordinator (LC) for
managing the nodes in that zone and one central global coordinator (GC) is chosen. The GC keeps
track of aggregated information of all the zones, while the LC has information about every node within
its area. Service Coordination is done by the participating LCs and the GC.
In HRPM, the entire network is divided into equal sized square cells. Within each cell there is
one access point (AP) and one rendezvous point (RP) for the entire network, for each of the service
group. Every node has the information about two different hash functions to find the position of the AP
and RP. The RP maintains information about all the service requests from different zones. The SP
requests for this information from the RP and constructs a multicast tree to deliver the service. HRPM
uses hierarchical routing based on greedy geographic forwarding algorithm.
We use a distributed directory based architecture, where we divide the entire geographic region
into zones as in [8]. We elect a Local coordinator (LC) for every geographic zone and a Global
Coordinator (GC) for entire region that is common to all the multicast groups, unlike the other schemes
that use one coordinator for every multicast group [15]. The LC is responsible for maintaining the
position information and service details of all the requestors and providers within a zone. The service is
registered by a provider with its LC and when the requestor needs a service it forwards the request to
its LC. The LC finds a suitable provider with the available service information or through the GC.
4. Leader Election for Service Provisioning
4.1. Role of a Leader Node in Service Provisioning
The role of a leader node is crucial in the different phases of service provisioning such as service
advertisement, service discovery, service coordination and service delivery. In directory based
architectures, the SPs register their services with the directory or leader node. In reactive service
discovery mode, a SR forwards the service request through the leader node. The service coordination is
a process of coordinating all the SPs, requestors and available services. This is done by the leader node
periodically and the service details are updated. Service delivery involves distributing a service to all
the requestors. Hierarchical service provisioning schemes divides the geographical area into various
zones and selects a leader node for every zone. The SP then constructs a tree from itself to the zone
leaders of all the requestors and then delivers the service to them.
4.2. Adaptive Leader Election
We use a vote based election algorithm [16] that elects a core node (LC or GC) dynamically. A node
calculates its eligibility factor (EF), based on the distance to the zone center, remaining battery power
and average speed.
1. When a new SP or requestor enters a zone, it sends a hello message to learn about the core
node to its neighbors.

10 Cynthia Jayapal and Sumathi Vembu
2. All neighbor nodes reply with a beacon message, using which a neighbor table is
constructed.
3. If there is no other core node, the new node announces itself as a leader and initiates an
election procedure to elect a new core node.
4. To elect a core node, initially all the nodes in transmission range to the center of zone
calculate their eligibility factor and send them to the leader.
5. The leader elects a node with the highest eligibility factor within a zone, as a new LC and
amongst all zones and within a radius of r/2 from the center of the geographical region, as a
new LC.
6. The node with the second highest ranking is stored as a backup node. If the core node fails or
gets overloaded, the backup node becomes the core node.
7. Whenever the eligibility factor of the core node becomes less than the threshold, the election
procedure is initiated to select a new core node.
8. All service records are transferred by the old core node to the new core node.
5. Adaptive Service Discovery Protocol (ASDP)
Table 1, illustrates the important tables used by us for service provisioning, with their functions,
descriptions and update details.
Table 1: List of tables maintained
S.
No. Table name Stored at Function Description Updated when Updated
by
1.
Service
Provider
(SP) Table
SP To store service
parameters.
Service Parameters:
Provider ID
Provider Location
Advertisement Seq. No.
Service Name,
Service ID,
Service Life Time
Change in service
parameter information. SP
2. Member
Table SP
To record details of
the service requests
received.
SR ID
SR’s LC ID
SR’s LC Location
Service ID
Service request is
received. SP
3.
Global Core
Node(GC)
Table
GC
To coordinate
activities of all
LC’s.
LC ID
LC Location
Service Parameters
LC moves, Change in
EF, Change in service
information.
LC
4.
Local Core
Node (LC)
Table
LC
To store details of
all the SP’s and
their services
within the zone.
Service Parameters SP moves, Change in
service information. SP
5.
Service
Requestor
(SR) Table
LC
To maintain service
requests within the
zone.
SR ID
SR location
SP ID
Service ID
Service request or reply
is made. SR / SP
6. Neighbor
Table Nodes To record details of
the Neighbors.
Neighbor ID
Neighbor Location
Flag (=1 if coordinator)
Change in location Neighbor
nodes
5.1. Service Registration and State Maintenance
All the SP in a zone wishing to provide any service, register their service with the respective LC by
sending the associated service parameters like advertisement sequence number, service name, service
ID, provider ID, location and service lifetime. The LC on receiving this information updates the LC

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 11
Table. On expiry of the lifetime of a service, the corresponding entry is deleted from the table. Any SP,
willing to extend its service, can re-register with the new lifetime before expiry. A SP, that wishes to
withdraw providing a service, directly informs its LC and the LC of all the SR zones. The LCs of all
the zones, advertise the LC Table contents, to the GC periodically and the GC updates its GC Table.
The LC Table and GC Table are used as a service cache by the coordinator nodes. The
coordinator nodes overhear the service advertisement and reply messages from the service requestors
and providers. The entry is refreshed on hearing an advertisement with a greater sequence number or
an entry is added if not found. The entries in the table are deleted after expiry of the service lifetime.
5.2. Service Discovery and Service Selection
Figure 1, illustrates the service discovery procedure. Any SR node which requires a service sends a
S
REQ
with its NID and service details to its LC. On receiving the S
REQ
the LC checks the LC Table if
there is a provider within the same zone, if so the S
REQ
is forwarded. Otherwise, the LC checks the SR
Table to see if there is already a request for the same service within the zone, if so the S
REP
is
forwarded directly to the requestor after updating the SR Table. If the SP is known to the SR’s LC, the
SREQ is directly forwarded to the SR’s LC. If a SP is not within the zone and is not known to the SR’s
LC, the LC forwards the S
REQ
to the GC. The GC checks the GC Table for a LC with a SP, providing
the requested service. If found the request is forwarded to the LC of the SP. Any coordinator node that
hears a SREQ forwards it directly to the provider, if it is known. On receiving a SREQ, the SP’s LC
gets the position of the SP from LC Table and forwards the request to the corresponding SP. The SP on
receiving the S
REQ
sends a S
REP
to the SR after updating its Member Table. If there is no match in the
GC Table, the S
REQ
is flooded.
In MANET, several SPs may provide the same service. The core node chooses the SP, before
forwarding the S
REQ
packet to the SP. The SP is chosen by considering the distance between the SR
and SP, the remaining lifetime of the service and the EF. This helps to prevent a single provider being
overloaded and also avoids redundant service delivery from all the providers like in [15].

12 Cynthia Jayapal and Sumathi Vembu
Figure 1: Service Discovery in ASDP
5.3. Location Update
Every node maintains the location of its neighbors and its LC. All LCs maintain the location of the GC
and the GCs maintain the location of all the LCs. Whenever a node enters a new zone it sends a
HELLO message to all its neighbors in the Neighbor Table. The neighbor nodes receiving this message
send back a BEACON message, with their location and LC information. If the sender finds that it has
moved to a new zone, it associates itself with the LC of that zone. If no Beacon is received, it assumes
that it is the only node in that zone and hence considers itself as LC of that zone. It sends an update
packet to the GC indicating its presence.
Every node stores and maintains details of its neighbors in its Neighbor Table. Location update
is done by a GC to all the LCs, LC to all the SRs and SPs in its zone, SR and SP to LC when it moves
more than 100m. A node updates its location to all its neighbors, if it moves beyond the transmission
range.

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 13
6. Performance Evaluation
6.1. Simulation Overview
The proposed scheme was implemented using NS2 [17] on LINUX platform. We have included the
energy model with initial energy set as 0.5 Joules, txPower as 0.3 watt and rxPoweras 0.6 watt. We
have assumed the MANET region as a square region of size 2400mX2400m. For virtual zone
construction, the zone-size is taken as 800m, which resulted in 9 different zones with Z
IDs
(0,0),(0,1),(0,2),(1,0),(1,1),(1,2),(2,0),(2,1), (2,2). The simulations were run for varying network size
with V
max
values ranging from 10 to 200 m/s and with different pause time. The simulation result was
gained by averaging over 10 runs. The nodes were randomly distributed using Random Waypoint
Model. We have assumed that at least 25 % of the total number of nodes in the network acts as SPs.
Each of these SP provides different service types with varying lifetimes. A service of a particular type
is uniquely identified by Service ID. A SP is assumed to provide a maximum of 5 different types of
services. SPs are simulated to provide similar service types. For evaluating the EF of a core node, we
gave 70% preference for remaining battery power, and 15% preference for both speed and distance to
center.
6.2. Simulation Results
We have studied the following metrics.
Service discovery success ratio - It is the ratio of the number of service request message
issued, to the corresponding hit message. In the simulation, the service request message was sent by the
SR and the hit message was sent by the SP. To study the effect of moving speed on Service discovery
success ratio, we varied the maximum speed of the nodes (V
max
) from 10 m/s to 200m/s. Figure 2,
shows that ASDP performance is better than SGSP and HRPM. We also infer that the change in
moving speed does affect the service discovery success ratio. The success ratio is more, when the
provider is within the zone and when it is known to the LC.
Figure 2: Effect of speed on service discovery success ratio
Control message overhead- It is the total number of control messages forwarded for service
discovery and hand offs. Figure 3, shows the effect of speed, pause time and group size on control
message overhead. It shows that the ASDP has less number of control message transfers because of the
more stable core nodes and less number of hand offs. ASDP and SGSP use similar provider selection
mechanism. Therefore the difference in overhead is only due to the stable core node of ASDP. The

14 Cynthia Jayapal and Sumathi Vembu
overhead in HRPM is more because, it sends control messages to all the available SPs and has frequent
handoffs.
Figure 3(a): Effect of speed on control message overhead
Figure 3(b): Effect of pause time on control message overhead

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 15
Figure 3(c): Effect of group size on control message overhead
Average discovery delay – It is the average time interval between a service request and service
reply. To study the effect of request load on the discovery delay, we varied the number of SRs and
observed that with increase in the number of requests the delay increased for all the three techniques.
Figure 4, shows that the discovery delay is more in HRPM as the hash value is computed to find the
AP and RP location and also because of the frequent change in core nodes. Performance of SGSP and
ASDP are similar under light load. As the load increased the average discovery delay of SGSP
increased linearly but there is marginal increase of delay for ASDP.
Figure 4: Effect of number of requestors on average discovery delay
Joining delay – It is the time difference between a node joining the group and receiving its first
data packet. To study the effect of the total number of nodes on the joining delay, we varied the total
number of nodes in the network from 30 through 50. It is observed from Figure 5, that SGSP and
HRPM has higher joining delay than ASDP and it increases with increase in number of nodes. We
infer this is due to factors like increased distance between the coordinator and the new node. In ASDP
the delay is almost stable owing to optimum distance between the coordinator and new node and the
SP is directly intimated about the request. In HRPM, before each delivery the SP has to get the
membership update from the RP and hence the joining delay is more.

16 Cynthia Jayapal and Sumathi Vembu
Figure 5: Effect of the number of nodes on joining delay
Election rate- It is the number of times the election procedure for coordinator selection is
invoked. To study the effect of the total number of nodes on the election rate, we varied the total
number of nodes in the network from 30 through 50. It is observed from Figure 6, that HRPM has the
highest number of elections because whenever a new node comes near the hashed location, the core
node changes and the changes are more when the mobility is more. In SGSP, the leader node is
consistent until it moves out of the zone but the nodes battery power is not taken into consideration. In
ASDP election happens only when EF of a node drops below the threshold limit. When the leader
changes are more the control overhead is also more.
Figure 6: Effect of speed on number of leader change
7. Conclusion
Our adaptive leader election and service discovery procedure, selects a core node that is reliable and
stable. This in turn reduces the number of hand offs and the control overhead involved in service
discovery. Our simulation results shows that the performance of ASDP is better than the other
algorithms that use node ID and hash value for leader election, in terms of success ratio, discovery
delay and control message overhead. The ASDP protocol can be enhanced by dynamic and cooperative
cache update mechanism. We also propose to advertise the popular service to all back bone nodes to
improve the discovery delay.

Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks 17
References
[1] Seyed Amin Hosseini Seno, Rahmat Budiarto and Tat-Chee Wan, 2007. “Survey and new
Approach in Service Discovery and Advertisement for Mobile Ad hoc Networks”,
International Journal of Computer Science and Network Security 7, pp. 275-284.
[2] Christopher, N., Ververidis and George, C., Polyzos, 2008. “Service Discovery for Mobile Ad
Hoc Networks: A Survey of Issues and Techniques”, Communications Surveys & Tutorials,
IEEE Publication 10, pp. 30-45.
[3] Adnan Noor Mian, Roberto Baldoni and Roberto Beraldi. 2009. "A Survey of Service
Discovery Protocols in Multihop Mobile Ad Hoc Networks", IEEE Pervasive Computing 8, pp.
66-74.
[4] Vincent Lenders, Martin May and Berhard Plattner, 2005. “Service discovery in mobile ad hoc
networks: A field theoretic approach”, Elsevier Journal on Pervasive and Mobile Computing 1,
pp. 343-370.
[5] S. A,. Hosseini-seno1, B., Pahlevanzadeh, T.C., wan, R., Budiart, and M.M., Kadhum, 2009.
“A Distributed Resource-Management Approach in Manets”, Journal of Engineering Science
and Technology 4, pp. 142 – 153.
[6] F.,Sailhan and V.,Ussarbtm, 2005. “Scalable service discovery for MANET”, IEEE conference
on Pervasive Computing and Communication, pp. 235 -244.
[7] Hassan Artail, Khaleel Wafiq Mershad and Hicham Hamze, 2008. “DSDM: A Distributed
Service Discovery Model for MANETs”, IEEE Transaction on Parallel and Distributed
Systems 19, pp.1224 – 1236.
[8] Xiaojing Xiang and Xin Wang, 2007. “A Scalable Geographic Service Provision Framework
for Mobile Ad Hoc Networks”, IEEE conference on Pervasive Computing and Communication,
pp. 151-160.
[9] Xiaojing Xiang, Xin Wang and Zehua Zhou, 2006. “An Efficient Geographic Multicast
Protocol for Mobile Ad hoc Networks”, IEEE World of Wireless, Mobile and Multimedia
Networks, pp. 73 -82.
[10] Honghui Luo and M., Barbeau, 2004. “Performance Evaluation of Service Discovery Strategies
in Ad Hoc Networks”, Second Annual Conference on Communication Networks and Services
Research, pp. 61 – 68.
[11] Alex Varshavsky, Bradley Reid and Eyal de Lara, 2005. “A cross layer approach to service
discovery and selection in MANETs”,
IEEE International Conference on Mobile Ad hoc and
Sensor Systems, pp.466.
[12] C., Dabrowski, K., Mills and J., Elder 2002. “Understanding consistency maintenance in
service discovery architectures during communication failure”, Third International Workshop
on Software and Performance, ACM, pp. 15 – 20.
[13] Sudarshan vasudevan, Jim Kurose and Don Towsley, 2004. “Design and Analysis of a Leader
Election Algorithm for Mobile Ad Hoc Networks”, IEEE International Conference on Network
Protocols, pp. 350 -360.
[14] Qi Dong and Donggang Liu, 2009. “Resilient Cluster Leader Election for Wireless Sensor
Networks”, Sixth Annual IEEE Communications Society Conference on Sensor, Mesh and Ad
Hoc Communications and Networks, pp. 1 – 9.
[15] Saumitra M.Das, Himabindu Pucha, and Y., Charlie Hu, 2008. “Distributed Hashing for
Scalable Multicast in Wireless Adhoc Networks”, IEEE Transactions on Parallel and
Distributed Systems 19, pp.347-362.
[16] Cynthia Jayapal, Sumathi Vembu and Arul Jothi Subramanium, 2010. “Adaptive Service
Provisioning for Mobile Ad Hoc Networks”, ICTACT Journal on Communication Technology
3, pp. 166-171.
[17] NS-2 Simulator, http://www.insi.edu/nsname/ns. Accessed 10 December 2008.