A Low Control Overhead Cluster Maintenance Scheme for Mobile Ad hoc NETworks (MANETs)

ACEEE International Journal on Network Security, Vol 1, No. 1, Jan 2010

A Low Control Overhead Cluster Maintenance
Scheme for
Mobile Ad hoc NETworks (MANETs)
Narendra Singh Yadav 1, Bhaskar P Deosarkar 2, and R.P.Yadav3
Department of Electronics & Communication Engineering,
Malaviya National Institute of Technology,
Jaipur, India
Email: {narensinghyadav@yahoo.com, bhaskar44_nanded@yahoo.co.in, rp_yadav@yahoo.com}

Abstract— Clustering is an important research area for dynamic topology, limited physical security, peer-to-peer
mobile ad hoc networks (MANETs) as it increases the nature, utilization of multihop relaying and dependency
capacity of network, reduces the routing overhead and on battery life. The major function of any data network is
makes the network more scalable in the presence of both to transport the data, from the intended source to the
high mobility and a large number of mobile nodes. In
clustering the clusterhead manage and store recent routing
desired destination, in a reliable way and in a minimum
information. However the frequent change of clusterhead time. But the error prone nature and limited range of the
leads to loss of routing information stored, changes the route shared wireless medium, absence of any fixed
between two nodes, affects the performance of the routing infrastructure, limited resources of the mobile nodes and
protocol and makes the cluster structure unstable. the dynamic topology impose certain restrictions on
Communication overhead in terms of exchanging messages establishing and maintaining the routes for such a data
is needed to elect a new clusterhead. The goal then would be delivery and make the task of routing and resource
to keep the clusterhead change as least as possible to make management difficult and challenging. Several protocols
cluster structure more stable, to prevent loss of routing and architectures [1] [2] are proposed in the literature for
information which in turn improve the performance of
routing protocol based on clustering. This can be achieved
performing this task which may broadly classified as flat
by an efficient cluster maintenance scheme. In this work, a and hierarchical based on the topological arrangement of
novel clustering algorithm, namely Incremental nodes assumed.
Maintenance Clustering Scheme (IMS) is proposed for To meet the expected demand of allowing the new
Mobile Ad Hoc Networks. The goals are yielding low nodes to join the network and existing nodes to leave the
number of clusterhead and clustermember changes, network, the architecture used should have sufficient
maintaining stable clusters, minimizing the number of scalability. But it is proved earlier that flat architecture is
clustering overhead. Through simulations the performance not as much scalable [3][4][5]. Clustering is proved to be
of IMS is compared with that of least cluster change (LCC) scalable and bandwidth efficient structure which basically
and maintenance scheme of Cluster Based Routing Protocol
(CBRP) in terms of the number of clusterhead changes,
forms a hierarchical arrangement of the nodes [6]. The
number of cluster-member changes and clustering overhead basic purpose behind clustering is to form and maintain a
by varying mobility and speed. The simulation results connected cluster structure. It consist of two phases- the
demonstrate the superiority of IMS over LCC and cluster formation which deals with building of cluster
maintenance scheme of CBRP. structure and the cluster maintenance that deals with
updating the cluster structure according to the changing
Index Terms—MANET, CBRP, cluster maintenance, control network topology and is quite important being related to
overhead, routing the performance of the given clustering algorithm.
The clusterhead manages and stores recent routing
I. INTRODUCTION information. Communication overhead in terms of
With the increase of small size information processing exchanging messages is needed to elect a new
devices, like laptop, pocket PC and PDA, the growing clusterhead. The frequent change of clusterhead makes
need to exchange digital information among people the cluster unstable due to loss of routing information
within a short communication range caused the which may cause the change in the route between two
emergence of Mobile Ad hoc NETworks (MANETs). nodes and hence affecting the performance of the routing
MANET can be defined as an autonomous system of protocol. The goal then would be to keep the clusterhead
mobile nodes connected by wireless links, in which the change as least as possible, to make cluster structure
nodes organize themselves arbitrarily and are free to more stable, to prevent loss of routing information which
move randomly. Some of their most interesting features in turn improve the performance of routing protocol. This
are the possibility of multi-hop communication, the lack can be achieved by an efficient cluster maintenance
of a fixed centralized infrastructure and capability of self- scheme. The basic idea behind this is to delay clusterhead
organization. These features made them attractive for change when two clusterheads are within the range of
battlefield, emergency operations such as search and each other so that the passing clusterhead moves out of
rescue in which the deployment of a fixed infrastructure the range of other clusterhead to avoid unnecessary
can be costly, risky, and time-consuming. MANETs are clusterhead change.
characterized by their distributed nature of operation, Accordingly various cluster maintenance schemes are
proposed in the literature in which when two clusterhead
1
© 2010 ACEEE
DOI: 01.ijns.01.01.01


come into range of each other one of the clusterhead always bear some specific attributes in its local area. But
leaves its role based on certain predefined criterion. the second case of re-clustering in LCC indicates that a
Particularly, in LCC [7] the criterion used may be Lowest single node’s movement may still invoke the complete
ID or Highest Connectivity based on the clustering cluster structure re-computation, and once this happens,
scheme used and the clusterhead change is delayed by a the large communication overhead for clustering may not
predefined time called as “contention period” in CBRP be avoided. And as and when two clusterhead come into
[8]. But these schemes may impose an unnecessary radio range of each other the clusterhead change is
clusterhead change when two clusterheads are just bought in. This is unwanted when two clusterheads are
passing each other. To avoid such unwanted clusterhead just passing each other and may be in each others range
change an Incremental cluster Maintenance Scheme for a short duration.
(IMS) is proposed in this paper.
B. Cluster maintenance in CBRP:
Rest of the paper is organized as follows. In section II,
In CBRP, as clusters are identified by their respective
LCC and cluster maintenance in CBRP are explained in
cluster heads it is desirable to have clusterhead changes
brief as are related to the scheme proposed in this paper.
as minimum as possible. For maintaining the cluster
The proposed Incremental Maintenance Scheme (IMS) is
following clusterhead change rules are imposed in CBRP,
presented in section III. Section IV explains the
as described in [8].
simulation parameters and performance metrics used.
Results are presented and discussed in Section V. Finally • A non-cluster head never challenges the status of
Section VI concludes this paper. an existing cluster head.
• When two cluster heads move next to each other
II. RELATED WORKS over an extended period of time (for
CONTENTION_PERIOD seconds), then only will
In this section previously proposed cluster one of them lose its role of cluster head.
maintenance schemes are discussed in brief. As a result, whenever a cluster head hears HELLO
messages from another cluster head indicating a bi-
A. Least clusterhead change [LCC]
In previous work, two simple criterions are used to directional link, it sets c_timer to expire in
form the clusters: One based on node ID and the other CONTENTION_PERIOD seconds. When c_timer
based on node degree. Specifically, lowest-ID nodes (LIC expires, it will check if it is still in contention with the
[9]) and maximum-degree nodes (HCC [10]) respectively other cluster head, by checking if the other cluster head
is still in its neighbor table. If so, it compares its own ID
are elected to be the clusterheads in cluster formation.
In LIC, the node with the lowest ID among its with that of the other cluster head's. The one with a
neighbors is elected as clusterhead. When a clusterhead smaller ID will continue to act as cluster head. The one
with a bigger ID gives up its role as cluster head and
finds a member with ID lower than its own ID then the
changes from C_HEAD to C_MEMBER in its
clusterhead is forced to handover the clusterhead role to
this node with lowest ID. subsequent HELLO messages. This might trigger
reorganization of other clusters. These rules guarantee
In HCC, the clustering is performed periodically to
check the “local highest node degree” attribute of a some sort cluster stability by delaying the clusterhead
clusterhead. When a clusterhead finds a member node change by CONTENTION_PERIOD upon coming of two
clusterheads in each others range. This avoids
with a higher degree, it is forced to relinquish its
clusterhead role. unnecessary clusterhead change if their passing time is
Both LIC and HCC mechanism involves frequent re- less than or equal to CONTENTION_PERIOD but if
clustering due to mobility of the nodes. To avoid frequent passing time is more the clusterhead change is forced.
re-clustering in these schemes an improvement is
suggested by LCC. III. PROPOSED INCREMENTAL MAINTENANCE SCHEME
In LCC the clustering algorithm is divided into two [IMS]
steps: cluster formation and cluster maintenance. The The cluster formation mechanism in IMS is based on
cluster formation simply follows LIC, i.e. initially mobile lowest ID clustering algorithm in which the node with
nodes with the lowest ID in their neighborhoods are lowest ID in neighborhood is elected as a clusterhead. In
chosen as clusterheads. Re-clustering is event-driven and the proposed scheme, when two clusterheads are with in
invoked in only two cases: range of each other, clusterhead change is delayed for
• When two clusterheads are within radio range of delay_period which is equal to Hello_interval initially. If
each other the clusterhead with lowest ID after delay_period both are again with in range of each
continues to work as clusterhead and forcing other other then delay_period is increased by Hello_interval.
to relinquish its role. A simple member node is not Delay_period is incremented by Hello_interval every
allowed to challenge the clusterhead even if it has time both clusterheads are within range of each other, till
an ID lower to clusterhead. delay_period is less than or equal to max_limit which is
• When a node cannot access any clusterhead, it obtained by dividing two times transmission range by
rebuilds the cluster structure for the network speed. If both are still with in range then the one with a
according to LIC. smaller ID will continue to act as a clusterhead and the
Hence, LCC significantly improves cluster stability by other one gives up its role as clusterhead as shown in
relinquishing the requirement that a clusterhead should following Fig. 1.

2
© 2010 ACEEE
DOI: 01.ijns.01.01.01


Receive Hello
from CH

Set delay timer = delay
delay timer expires
period

Is still in contention with
other CH?

YES

NO

Is
my_CH_ID < NO Is
Other CH_ID Delay period ≤ Max.
YES Limit

Send triggered Hello
as CH NO
YES
Send triggered
hello as CM

delay period += hello
interval

Figure 1. Flowchart depicting the clusterhead change in proposed scheme

IV. SIMULATION PARAMETERS AND PERFORMANCE scenario are simulated for 300 seconds to collect the
METRIC desired data at steady state to obtain statistically
confident averages.
LCC, cluster maintenance scheme in CBRP and the TABLE 1
proposed IMS are implemented in NS2 [11]. It is an SIMULATION PARAMETERS
object-oriented, discrete event driven network simulator
developed at UC Berkely written in C++ and OTcl and is Parameters Stage I Stage II
Number of mobile nodes, 150
particularly popular in the ad hoc networking research N
community. In this simulation study the source- Simulation Area 2000m x 500m
destination pairs are spread randomly over the network. Simulation Time 300s
The node movement generator of NS-2 is used to Pause time for mobile 0, 60, 120, 180, 100 s
generate the different node movement scenarios. The nodes 240 and 300s
Max. speed for mobile 10m/s 5, 10, 15, 20
node movement is assumed to follow the random way nodes and 25 m/s
point model. The movement generator takes the number Transmission range for 250m
of nodes, pause time, maximum speed, field mobile nodes
configuration and simulation time as input parameters. Performance Metrics: The metrics considered for
The propagation model used is two ray ground [12]. evaluations are the number of clusterhead change, the
Simulations consist of two stages. In stage1 simulations number of cluster member change and clustering
are carried out by varying the mobility (pause time) and overhead by varying pause time and speed.
in stage2 by varying the node speed. The simulation The number of clusterhead change is the total number
parameters used are listed in Table1. Five runs of each of clusterhead changes during the whole simulation run
3
© 2010 ACEEE
DOI: 01.ijns.01.01.01


time. A small value of clusterhead change reflects the
stability of the cluster structure. LCC CBRP IMS
The number of cluster member change is the number 500

# of cluster member changes
of mobile nodes that switch to another clusterhead during
400
the simulation run time.
Clustering overhead is the number of clustering 300
messages sent by each node in cluster formation and
cluster maintenance operation. It is an important measure 200
for the scalability of a protocol.
100
V. RESULTS AND DISCUSSION
0
Fig.2 - Fig. 4 shows the performance of IMS, CBRP 0 60 120 180 240 300
and LCC in terms of number of clusterhead changes, Pause time (sec)
number of cluster member changes and clustering
overhead as function of pause time and performance with Figure 3 Number of cluster member change vs. pause time
respect to speed is shown in Fig. 5 – Fig. 7 The pause
time is varied from 0 sec to 300 sec in steps of 60 sec and Fig. 3 shows the number of cluster member changes as
number of clusterhead changes, cluster member changes a function of pause time. In LCC, CBRP and IMS all as
and clustering overhead is observed. For observing the pause time increases the required number of cluster
effect of speed the node speed is varied from 5 m/s to 25 member changes are very low. From figure it is clear that
m/s in steps of 5 m/s and the performance metric is IMS performs better to LCC and CBRP both. At highest
evaluated. mobility when the nodes are continuously moving the
Fig. 2 shows the number of (#) clusterhead changes as number of cluster member changes required in IMS are
a function of pause time. In LCC, CBRP and IMS all as approximately one ninth of that of LCC and fourth of
pause time increases the required number of clusterhead CBRP. Pause time upto 180 sec CBRP scheme performs
changes are very low. From figure it is clear that IMS better to LCC after which both perform similarly whereas
performs better to LCC and CBRP both. At highest IMS outperforms both of them. This change in behavior
mobility when the nodes are continuously moving the can be reasoned to the number of clusterhead changes
number of clusterhead changes required in IMS are required in respective schemes. The less number of
approximately one third that of LCC and half of CBRP. clusterhead changes indicate the reduced number of re-
Pause time upto 180 sec CBRP scheme performs slightly affiliations in IMS as compared to LCC and CBRP
better to LCC after which both perform similarly whereas schemes.
IMS outperforms both of them. This difference in
performance is due to the clusterhead change delay
strategy used in CBRP and IMS. In CBRP the delay is of 26500 LCC CBRP IMS
26400
the time equal to CONTENTION_PERIOD whereas in
Clustering overhead

26300
IMS it is a function of speed and transmission range.
26200
26100
26000
90
LCC CBRP IMS
25900
# of clusterhead changes

80 25800
70 25700
60 25600
50 0 60 120 180 240 300
40 Pause time (sec)
30
20 Figure 4 clustering overhead vs. pause time
10
0 Fig. 4 shows the clustering overhead, in LCC, CBRP
0 60 120 180 240 300 and IMS, as a function of pause time. From figure it can
Pause time (sec) be observed that the clustering overhead required in IMS
are comparatively less than LCC and CBRP. In IMS the
Figure 2 Number of clusterhead change vs. pause time less number of clusterhead changes triggers less number
of cluster member re-affiliations which are the main
sources of control overhead.
Fig.5 below shows variation in clusterhead changes
with respect to variation in speed. In all three schemes
LCC, CBRP and IMS, the number of clusterhead changes
increases with increase in speed. The number of
clusterhead changes in IMS at all the speeds between 5
4
© 2010 ACEEE
DOI: 01.ijns.01.01.01


m/s and 25 m/s are very less as compared to LCC and observed that the clustering overhead required in IMS are
CBRP. The sudden increase in number of clusterhead comparatively less than LCC and CBRP. In IMS the less
changes at 20m/s may be due to simulation restrictions. number of clusterhead changes triggers less number of
cluster member re-affiliations which are the main sources
LCC CBRP IMS of control overhead.
60
# of clusterhead change

50 VI. CONCLUSION
40 Since the stability of cluster in a cluster based Mobile
30 Ad hoc networks affects the performance of protocols
such as scheduling, routing and signaling, a clustering
20
maintenance scheme, named IMS, had been proposed and
10 studied in this paper. The basic idea of this scheme is to
0 delay clusterhead change when two clusterheads are
5 10 15 20 25 within transmission range of each other to avoid
Speed (m/s) unnecessary clusterhead change. Simulation results
show that IMS is better cluster maintenance scheme as
Figure 5 Number of clusterhead change vs. speed compared to LCC and the scheme used in CBRP in terms
of number of clusterhead changes, number of cluster
Fig. 6 shows variation in cluster member changes with member changes and clustering overhead. This is due to
respect to variation in speed. In all three schemes LCC, avoiding the unnecessary clusterhead changes. In
CBRP and IMS, the number of cluster member changes conclusion, IMS successfully fulfills its aim of providing
increases with increase in speed. The less number of a stable cluster structure for MANETs
clusterhead changes in IMS causes less cluster member
re-affiliations and hence the number of cluster member REFERENCES
changes required in IMS are very less than that of LCC [1] M. Abolhasan, T. Wysocki, A. Dutkiewicz, “A review of
and CBRP. routing protocols for mobile ad hoc networks”, Elsevier
Science, Journal of Ad Hoc Networks, Vol.2, 2004, pp.1-
LCC CBRP IMS 22
[2] E. M. Royer and C. K. Toh, “A review of current routing
# of cluster-member changes

350
protocols for ad hoc mobile wireless networks,” IEEE
300 Personal Communications magazine, April 1999, pp. 46–
250 55.
200 [3] P. Gupta and P. R. Kumar, “The Capacity of Wireless
Networks,” IEEE Trans. Info. Theory, vol- IT 46.2, Mar.
150 2000, pp. 388-404.
100 [4] X. Y. Hong, K. X. Xu and M. Gerla, “Scalable Routing
50 Protocols for Mobile Ad Hoc Networks,” IEEE Network,
July-Aug. 2002, pp. 11-21.
0 [5] X. Y. Hong, K. X. Xu and M. Gerla, “An Ad Hoc Network
5 10 15 20 25 with Mobile Backbones,” Proc. IEEE ICC 2002, vol. 5,
Speed (m/s) Apr.-May 2002, pp. 3138-43
[6] E. M. Belding-Royer, “Hierarchical Routing in Ad Hoc
Figure 6 Number of cluster member change vs. speed Mobile Networks,” Wireless Commun. And Mobile
Comp., vol. 2, no. 5, 2002, pp. 515-32.
LCC CBRP IMS [7] C.-C. Chiang, H.-K. Wu, W. Liu and M. Gerla, “Routing
26400 in Clustered Multihop, Mobile Wireless Networks with
Fading Channel,” in Proc. IEEE SICON’97, 1997.
26300
Clustering overhead

[8] Mingliang Jiang, Jinyang Li and Y.C.Tay, “Cluster Based
26200 Routing Protocol”, August 1999 IETF Draft.
http://www.ietf.org/internet-drafts/draft-ietf-manetcbrp-
26100 spec-01.txt
26000 [9] A. Ephremides, J. E. Wieselthier, and D. J. Baker, “A
Design Concept for Reliable Mobile Radio Networks with
25900 Frequency Hopping Signaling,” in Proc. IEEE, vol. 75,
1987, pp. 56–73.
25800
[10] M. Gerla and J. T. Tsai, “Multiuser, Mobile, Multimedia
5 10 15 20 25 Radio Network,” Wireless Networks, vol. 1, Oct. 1995, pp.
Speed (m/s) 255–65.
[11] K. Fall and K. Vardhan, The Network Simulator (ns-2).
Figure 7 clustering overhead vs. speed Available: http://www.isi.edu/nsnam/ns
[12] T. S. Rappaport, Wireless Communications, Principles &
Fig. 7 shows the clustering overhead, in LCC, CBRP Practices. Prentice Hall, 1996, ch. 3, pp. 70-74.
and IMS, as a function of speed. From figure it can be

5
© 2010 ACEEE
DOI: 01.ijns.01.01.01

A Low Control Overhead Cluster Maintenance Scheme for Mobile Ad hoc NETworks (MANETs)

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to A Low Control Overhead Cluster Maintenance Scheme for Mobile Ad hoc NETworks (MANETs) (20)

More from IDES Editor (20)

Recently uploaded (20)

A Low Control Overhead Cluster Maintenance Scheme for Mobile Ad hoc NETworks (MANETs)