Dynamic Cluster Management In Ad hoc Networks Puneet Sethi Ubinetics India Pvt. Ltd., Bangalore - 560 027 e-mail: Puneet.Sethi@ubinetics.co.in Gautam Barua Department of Computer Science and Engineering, Indian Institute of Technology Guwahati, Guwahati - 781039 e-mail: gb@iitg.ernet.in Abstract—Clustering in ad hoc networks provides significant support for implementation of QoS and security, by overcoming inherent network defi- ciencies (like lack of infrastructure etc.). In a clustered network architecture, the ad hoc network is divided into groups of nodes called clusters. The clus- ters are dynamically maintained and reconfigured using specific protocols and algorithms. In this paper, we describe a clustering technique which has been designed as a part of the CRESQ([12]) routing algorithm for ad hoc networks. CRESQ’s clustering is segregated into two parts, the initial clustering and the cluster management. Initial Clustering creates the clusters during the network formation and cluster management maintains the clusters throughout the net- work lifetime. The main focus of this paper is the cluster management algo- rithm which is designed to be distributed in nature, adaptive to practicalities such as loss of packets and tolerant to mobilities during the algorithm exe- cution. The algorithm makes no assumptions about the availability of infor- mation concerning transmission range, node mobilities etc. This paper also discusses certain performance metrics for judging the efficiency of clustering in ad hoc networks. We also present simulation results for the cluster man- agement algorithm, along with a comparison with some existing algorithms. Results showing a comparison of CRESQ with existing protocols in terms of packet drop ratio and routing overhead are also provided. Keywords: clusters, routing, QoS. I. Introduction An ad hoc network is made up of mobile nodes with wireless in- terfaces, which communicate with each other over multihop paths, in the absence of any fixed infrastructure. Each mobile node in the network also acts as a router for the network traffic. The main advantages of such networks are rapid deployment and dynamic reconfiguration, which make them the right candidate for mili- tary applications, rescue operations and disaster recovery. Ow- ing to their ability to provide a quick and cheap communication link without the need of wired infrastructure, they help in extend- ing the limits of ubiquitous computing and information access for their users. Of late, there has been a need to support QoS and security in ad hoc networks owing to their varied usage scenar- ios and convergence of different applications’ traffic. Clustering provides a solution to support QoS in ad hoc networks, which are usually constrained by low computational and bandwidth capac- ity of nodes, mobility of intermediate nodes in an established path, and absence of routing infrastructure. In a clustered network architecture, the whole network is di- vided into self-managed groups of nodes called clusters. All the nodes within a cluster are at most two hops away from each other. These clusters continually adapt themselves to the changing net- work topology and new cluster configurations that are feasible with the current network topology, are created dynamically. Mas- ter (or Clusterhead) is the node which is only one hop away from all the other nodes in the cluster, and carries certain extra respon- sibilities. Much of the existing literature([11], [5], [15], [14], [10]) in this field focuses on the efficient partitioning of a network. The lack of an efficient complete and practical solution for ad hoc net- works, particularly concerning the integration of clustering with routing, motivated a study in this area. Clustering in an ad hoc network can be divided into two phases: cluster initialization i.e. the creation of the clusters and cluster management i.e. the main- tenance of clusters during the network lifetime. Most of the liter- ature only focuses on the efficiency of the first phase i.e. cluster initialization. The cluster maintenance in [5] is handled by up- dates; the mobile node checks for the presence of its clusterhead and cluster-partners before making an update, which generates a considerable overhead. The schemes described in [14], [15], [10] assume the availability of, information about node’s mobility and transmission range, feedback about the power of received signal, and guaranteed delivery of packets. Such assumptions are very difficult to ensure in a practical implementation. The cluster man- agement in [10] also assumes that the mobile node is aware of its trajectory and its cluster head can predict its future cluster- head. A distributed network partitioning technique is described in ACRPN([7]), which is also used as the initial clustering algorithm in our simulations. ACRPN also describes a quorum system to avoid degradation in performance during frequent cluster changes. Most of the existing schemes fail to deal with issues like main- taining consistent cluster information in the nodes, packet drops, overloading of nodes’ resources (due to unbounded cluster size), node movements during algorithm execution. In this paper we present a clustering technique which is a part of the CRESQ([12]) protocol. We attempt to provide a clustering solution which is distributed in nature, handles the cluster man- agement by taking into account practicalities like packet losses etc. and easily integrates with a routing module. Cluster based Routing for End-to-end Security and Quality of service satisfac- tion i.e. CRESQ, is a cluster based source routing protocol, which provides adequate support for QoS and security. The clustering in CRESQ is handled by two algorithms, initial clustering algorithm and cluster management algorithm. The initial clustering algo- rithm creates the clusters during the formation of the network and the cluster management algorithm maintains the clusters mainly using periodic transmission of HELLO packets. In this paper, we focus on the cluster management algorithm and its evaluation, and other details of CRESQ can be found in [12]. In the next section, we describe the clustering algorithm. In Section III, we discuss certain performance metrics, for clustering in ad hoc networks. In Section IV, we present the results of simulations and compar- isons of the cluster management algorithm with some existing al- gorithms. In Section V, a conclusion of the work is provided. II. Clustering Methodology The proposed cluster management algorithm was designed as a part of the CRESQ ad hoc routing protocol. Clustering in the CRESQ protocol helps in, reducing routing overhead, providing QoS guarantees and reducing QoS re-establishment delay during