99 CHAPTER 5 AN AODV-BASED CLUSTERING APPROACH FOR EFFICIENT ROUTING 5.1 INTRODUCTION Dynamic network topology and limited system resources characterize mobile ad hoc networking. Many routing protocols were proposed for routing in MANETs (Chakeres and Belding-Royer 2004; Jiang et al 1998). In MANETs, performance decreases when network’s size exceeds a certain threshold, resulting in many routing algorithms performing only when network’s size is small. To overcome bandwidth and battery power limitations and reduce routing overhead, it is mandatory to make network organization smaller and manageable (Garcia et al 2003). A clustering architecture provides MANET environments with three features: network scalability, fault tolerance and communication overhead reduction. Many clustering algorithms have geographical regions as clusters or form new ones unnecessarily. (Chiang et al 19997; Lin and Gerla 1997). An algorithm by Chatterjee et al (2000) creates necessary clusters without using routing protocol maintained information. It has been proved that cluster architecture ensures basic performance achievement in MANETs with many mobile terminals. A cluster structure provides at least three benefits, as an effective topology control measure. First, it facilitates spatial resources reuse to increase system capacity. With the non-overlapping multi-cluster structure, two clusters can use the same frequency/code set if they are not neighbours. Also, a cluster
22
Embed
CHAPTER 5 AN AODV-BASED CLUSTERING APPROACH FOR EFFICIENT …shodhganga.inflibnet.ac.in/bitstream/10603/24649/10/10_chapter5.pdf · AN AODV-BASED CLUSTERING APPROACH FOR EFFICIENT
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
99
CHAPTER 5
AN AODV-BASED CLUSTERING APPROACH FOR
EFFICIENT ROUTING
5.1 INTRODUCTION
Dynamic network topology and limited system resources
characterize mobile ad hoc networking. Many routing protocols were
proposed for routing in MANETs (Chakeres and Belding-Royer 2004; Jiang
et al 1998). In MANETs, performance decreases when network’s size
exceeds a certain threshold, resulting in many routing algorithms performing
only when network’s size is small. To overcome bandwidth and battery power
limitations and reduce routing overhead, it is mandatory to make network
organization smaller and manageable (Garcia et al 2003).
A clustering architecture provides MANET environments with
three features: network scalability, fault tolerance and communication
overhead reduction. Many clustering algorithms have geographical regions as
clusters or form new ones unnecessarily. (Chiang et al 19997; Lin and Gerla
1997). An algorithm by Chatterjee et al (2000) creates necessary clusters
without using routing protocol maintained information.
It has been proved that cluster architecture ensures basic
performance achievement in MANETs with many mobile terminals. A cluster
structure provides at least three benefits, as an effective topology control
measure. First, it facilitates spatial resources reuse to increase system
capacity. With the non-overlapping multi-cluster structure, two clusters can
use the same frequency/code set if they are not neighbours. Also, a cluster
100
coordinates transmission events better with a special mobile node, with a
cluster head, inside. This saves retransmission resources resulting in lower
transmission collisions. The second benefit is in routing, as a cluster head set
and cluster gateways normally are the virtual backbone for inter-cluster
routing which in turn restricts creation and spread of routing information.
Last, a cluster structure shrinks an ad hoc network and makes it more stable
with every mobile terminal. When a mobile node changes its cluster
attachment, the information is updated only in those nodes in corresponding
clusters. Thus, local changes do not need updation by the whole network as
this greatly reduces information stored by each mobile node.
5.2 CLUSTERING
Clustering technique is not a routing protocol; it aggregates nodes
into groups to ensure easier network management. An ad hoc network’s
cluster organization is impossible to be achieved in fixed infrastructure,
offline. Clustering scheme implementation ensures better protocol for Media
Access Control (MAC) through improvement of spatial reuse, throughput,
scalability and power consumption.
MANETs have multiple benefits from clustering. To begin with,
cluster structure ensures MANETs scalability and load balancing, increases
system capacity facilitating spatial resources reuse. In a multi cluster
structure, two non-overlapping clusters can have similar frequency if they are
not adjacent. It also elects a mobile host with special features called cluster
head for enhanced transmission activities coordination which in turn
alleviates collision of mobile nodes, saving both energy and resources.
Second, it limits creation and spreading routing information through
formation of a inter-cluster routing backbone consisting of cluster heads and
cluster gateways. Lastly, it provides a stable and smaller vision of ad hoc
network when a mobile node changes its attaching cluster so that the whole
101
network need not be aware of local changes in a node set. Only mobile nodes
in a corresponding cluster reflect such changes. Hence each node stores and
processes a fraction of total network routing information saving resources.
Clustering aims to be connected and nodes play various roles in
clustering techniques. The three node types are defined as follows:
1. Ordinary nodes
2. Cluster head
3. Cluster gateways
5.2.1 Cost Involved with Clustering
Construction and maintenance of a cluster structure requires
additional cost compared to flat structure MANETs. Cost of clustering
analysis is carried out either quantitatively or qualitatively to outline both
clustering techniques benefits and drawbacks. Clustering associated cost is
explained below:
1. Due to Frequent network topology changes cluster relatedinformation varies drastically in a dynamically changingcluster structure. Transfer of message packets subsequentlyeats up massive bandwidth, depleting mobile nodes energyand further disabling upper layer applications that are unableto be implemented with meagre resources.
2. Re-clustering occurs in some clustering schemes due toabrupt local instance like mobile nodes moving to otherclusters or due to the demise of a mobile node or shuttingdown of cluster heads, resulting in other cluster heads re-election. This is re clustering ripple effect which stimulates are-clustering effect over the full network.
102
3. Clustering stages are divided into two: cluster formation and
maintenance. The first stage assumes that mobile nodes are
static. Every mobile node obtains information from
neighbouring nodes in a specific period which might prove
impractical in a real time scenario.
5.3 CLUSTER MANAGEMENT
Clustering architecture provide network scalability, fault tolerance
resulting in increased use of network resources. It could be used for resource
management, routing and location management to lower communication and
computational overheads. This section discusses cluster formation and
maintenance.
5.3.1 Clustering Algorithm Design Goal
It is intended to integrate clustering with routing functions. The
design aims of the clustering scheme include:
An algorithm using a routing protocol’s control messages to
form clusters with limited overhead.
An algorithm operating in localized and distributed manners
and intertwining with nodes using only AODV.
The algorithm incurring limited cluster formation/
maintenance overhead and supporting formation of on-
demand clusters.
The algorithm minimizing network-wide flooding and being
scalable.
103
The proposed scheme constructs/updates clustering architecture
when clusters alone service is required. The on-demand nature of AODV’s is
utilized for the scheme. Nodes participating in clustering are known from CHs
maintained topological information and individual nodes.
5.3.2 Cluster Formation
The aim of clustering is to ensure efficient use of network
resources, enhance availability, reduce overheads and give scalable
architecture (Jiang et al 1998; Garcai et al 2003). The clustering algorithm
choice affects clusters stability. The proposed work divides a network into
several two-hop clusters where in each cluster a node can play one of five