Optimal Cluster Head Selection for Energy Minimization in Wireless Networks

T.Saranya, A.Karthikeyan
Department of Information Technology Bannari Amman Institute of Technology, Erode, India {saranya9486, yourskeyan}@gmail.com Abstract. The Mobile ad-hoc networks are autonomous systems of mobile nodes forming in the
absence of any centralized support. It has no fixed infrastructure. It is very difficult for routing in a challenging complex distributed system. The main aim of this paper is to improve performance of QOS by providing fault tolerance with efficient utilization of energy and satisfying all the time constrains. Performance is compared for both FDCB and EFDCB. In Fully Distributed ClusterBased (EFDCB) routing, it is very difficult to route the packet when link failure occur through the network. It may increase the time latency to create a new path. In Extended Fully Distributed Cluster-Based (EFDCB) routing, Cluster heads loses more energy as these are responsible for monitoring and updating the cluster QOS table as well as handling connection failures. By using HEFC algorithm, cluster heads in each cluster are changed after some time and saving the energy of the network with reduced time latency.

Keywords: Fault-tolerant distributed routing, mobile computing, wireless networks, FDCB

routing, EFDCB routing and HEFC algorithm. .

INTRODUCTION
Mobility has become very important, as people want to communicate anytime from and to anywhere. In the areas where there is little or no infrastructure is available or the existing wireless infrastructure is expensive and inconvenient to use, Mobile Ad hoc NETworks, called MANETs, are becoming useful. They are going to become integral part of next generation mobile services. A MANET is a collection of wireless nodes that can dynamically form a network to exchange information without using any pre-existing fixed network infrastructure. Mobile ad hoc networks (MANETs) require QoS capabilities that provide fault tolerance and fast recovery. Ad hoc wireless network is a collection of wireless mobile nodes forming a temporary network without any centralized administration. High churn or node mobility can cause QoS requirements to become unachievable. Excessive node mobility can lead to topology changes before network updates can propagate.

INTRODUCTION
This paper addresses stability and recoverability, two main issues in routing QoS traffic in mobile networks. The first issue is stability. With most ad hoc wireless networks that support QoS, each node acts as a router. In many distributed reactive routing schemes, if a node does not know the QoS parameters of its neighbors it broadcasts the route request packet and the neighboring nodes share their QoS parameters using broadcast packets. The broadcast packets used to discover the QoS parameters of nodes neighbors and negotiate QoS paths can flood the network. A clustered approach can lower this communication overhead to more scalable levels by limiting inter cluster control communication to gateway nodes. The second issue is minimizing the QoS impact due to network failures. If a supporting node fails when traffic is routed through multiple hops then, in the worst case, the connection must be rerouted from the source. This global fault-recovery method requires that the source renegotiate a new QoS path, which is costly in computation and communication. By contrast, if a protocol allows intermediate nodes to repair connections locally then the associated connections will likely only suffer minor disruption. Local repair can make the difference in meeting time-sensitive deadlines. This paper presents the key features, definitions, and assumptions of the extended fully distributed clusterbased (EFDCB) routing protocol, which is a fault-tolerant extension to FDCB. The main aim of this paper is to improve the performance of QoS with efficient usage of energy and limited time consumption using Higher Energy First Cluster (HEFC) algorithm.

RELATED WORK

A. Hierarchical Clustering
A typical Hierarchical Clustering is comprised of several cluster head nodes, and regular nodes. For administrative purposes, the operation of a Hierarchical Clustering is divided into rounds in which

nodes are grouped into clusters. Each round consists of three phases: cluster head selection (CHS), cluster formation (CFM), and data communication (DCM). The deterministic behaviors of a Hierarchical Clustering are typically characterized by the above three phases. However, the CHS phase plays the most dominant role with respect to the optimality and predictability of the entire network operation. A smart cluster head selection strategy can significantly reduce energy consumption, which in turn prolongs the network lifetime. Furthermore, a rule-based cluster head selection strategy can make the network lifetime more predictable. B. Routing Protocols 1) Proactive Routing Protocols: They are also called table driven protocols. It maintains routing table using the routing information learnt from neighbors on periodic basis. Main characteristics of these protocols include: distributed, shortest-path protocols, maintain routes between every host pair at all times, based on Periodic updates of routing table and high routing overhead and consumes more bandwidth. 2) Reactive Routing Protocols: They are also called demand driven protocols that find path as and when required. They maintain information about the active routes only. They performs route discovery phase before data transmission by flooding route request packet and destination node reply with route reply packet. A separate route maintenance procedure is required in case of route failure. Main Characteristics of these routing protocols are determining routes as and when required, less routing overhead, source initiated route discovery and more route discovery delay. C. The Dynamic Clustering Algorithm The main goal of this technique is to provide scalability, adaptability and autonomy through a hybrid approach and to achieve strong connectivity in the resulting network. In order to provide scalability and to enhance the availability (by providing the service locally), the network is partitioned into a number of non-overlapping groups called clusters. In the conventional approach, each cluster has exactly one distinguished node, designated as cluster head, which is responsible for organizing and establishing the cluster. Cluster-head election algorithms are used to elect these cluster heads. The main bottleneck in this approach is the single point of failure (if the cluster head crashes, the entire QoS parameter table information will be lost), which forces the above procedure to be repeated for the construction of the QoS parameter table and subsequent election of the cluster head. To overcome the above problems, a fully distributed architecture is proposed. Clusters are created using a cluster creation algorithm and each cluster member maintains a QoS parameter table (about its cluster members) and a gateway table. Gateway nodes manage the communication with adjacent clusters. Clustering is also used by some routing protocols in ad hoc networks. Routing is typically divided into two types: routing within the cluster (intra-cluster routing) and routing between different clusters (inter-cluster routing). D. Localized Quality-Of-Service Routing

1) Local Proportional Sticky Routing (PSR): The PSR scheme can be viewed to operate in two stages: proportional flow routing and computation of flow proportions. The algorithm proceeds in cycles of variable lengths. During each cycle, any incoming connection request can be routed among paths selected from a set of eligible paths, which initially may include all candidate paths. A candidate path is ineligible depending on the maximum permissible flow blocking parameter, which determines how many times this candidate path can block a connection request before it becomes ineligible. When all candidate paths become ineligible a cycle ends and all parameters are reset to start the next cycle. An eligible path is finally selected depending on its flow proportion. The larger flow proportions, the larger chances for selection. When a cycle is complete, a new flow proportion is found for each path based on blocking probabilities. 2) Credit-Based Routing (CBR): CBR performs routing using crediting scheme that rewards a path upon flow acceptance and penalizes it upon flow rejection. CBR requires every node to maintain a predetermined set of candidate paths R to each possible destination. It has been shown that routing algorithms that prefer short paths usually outperform algorithms that do not take path length into consideration. 3) Quality-Based Routing (QBR): The Quality Based Routing (QBR) introduces a heuristic algorithm to decide which path should be used to route traffic from source to destination based on the value of the QoS metric used. QBR monitors the current residual bandwidth for each path on its bottleneck links, and incorporates its values into simple average path qualities. Like CBR, QBR rewards a path which has a good quality when flow is successful and punishes it on when flow fails.

SYSTEM ANALYSIS

A. Fully Distributed Cluster-Based (FDCB) Routing This algorithm maintains the QoS parameter table at each cluster member with which every node finds all the feasible paths to all other nodes in the network. Nodes autonomously form clusters by exchanging the bid-request packets. Each cluster member maintains a QoS parameter table. Cluster creation technique used in this paper is simple and self terminating. Cluster based routing protocol limits the amount of routing information stored and maintained at individual hosts, thus reducing the memory requirements. Node mobility related events are handled locally within the clusters. Hence, farreaching effects of topological changes are minimal. Conventional cluster based algorithms use cluster head election algorithms to elect the cluster head. The cluster head maintains the QoS parameter table. If the cluster head crashes, all routing information will be lost. Network traffic is also centred towards the cluster head. Such networks are susceptible to single point failures. In the fully distributed architecture, all cluster members maintain the QoS parameter table, and is free from single point failure. Cluster maintenance is lightweight compared to cluster head election process. Traffic is regulated according to the routing table data. Using these QoS parameters, our algorithm constructs an optimal multicast tree satisfying the QoS requirements. Analysis shows that the performance of this algorithm is not adversely affected by the increase in network size, but it adds initial latency for the route discovery. With FDCB, scalability issues in centralized routing are circumvented. The FDCB method is similar to hierarchical routing in that each cluster node only maintains QoS information for other cluster members, a fraction of the network. Thus, an increase in nodes should not significantly increase memory or runtime. Further, since global network state is shared and maintained by all, the communication overhead is greatly reduced. In FDCB, if a flows source and destination are not in the same cluster, the source sends a route request packet to the gateway node, which forwards it to adjacent cluster(s). As long as the intermediate gateway nodes and links can support the requested QoS constraints, this process is repeated until the destination is found. The discovered path is sent back to the source and the resource reservation made. The distributed nature of FDCB allows it to avoid unmanageable shared global state. FDCBs distributed routing adds initial latency for the route discovery. B. Extended Fully Distributed Cluster-Based (EFDCB) Routing EFDCB reduces the impact of a connection failure since the cluster-head has complete cluster connectivity awareness. The result is efficient fault-tolerant QoS route maintenance for MANETs. EFDCB is primarily message driven, as is FDCB. The procedure executed by a node depends on the message it receives. Several message types are exchanged between nodes.

Fig1. Clustered ad hoc network

Extended Fully Distributed Cluster-Based EFDCB algorithm is a fault-tolerant extension to FDCB. It is targeted toward routing QoS packets in challenging MANET environments where links can break often and without warning. In these environments, a routing algorithm needs a contingency plan for link breakages. EFDCB provides QoS disruption mitigation. When EFDCB is successful, packets are delivered such that the applications dependent upon the network are fully functional. Conversely, if the protocol fails then the dependent applications could suffer lengthy QoS disruptions since the source will have to resort to rerouting. Technique is to repair the broken path at the failed link by shifting traffic to a neighboring node and then routing around the breaking point. This method avoids the costly process of rerouting the traffic from the source. EFDCB uses a cluster-head model to mitigate connection failures. This model employs cluster state knowledge sharing with to make the cluster-head aware of supported QoS connections in the cluster. Cluster state knowledge is shared with all cluster members for use in the event of a clusterhead failure. CFSR allows each node in the EFDCB network to be aware of the complete network state with low bandwidth impact. EFDCB is different from FDCB since link failures are handled locally instead of rerouting traffic from the source and QoS traffic is routed with lower packet transmission delays. The EFDCB faulttolerant method differs from Chen and Nahrstedts work in that a clustered approach is used to help avoid their repair methods limitations. EFDCB does not require that the predecessor of a failed node be capable of reaching the failed nodes successor. EFDCB presents an innovative solution to faulttolerant QoS supporting MANETs.FDCB uses an on-demand reactive routing scheme, but EFDCB adopts a more proactive approach. The QoS routing scheme used by EFDCB is Clustered Fisheye State Routing (CFSR). CFSR proposes a clustering framework to reduce redundant broadcast routing control messages. In CFSR, cluster-heads and gateways execute the original FSR protocol to send link state updates about the cluster, while ordinary nodes only send link state about themselves. This limits the messages from much of the network (ordinary nodes). The result is lower overhead. Assuming combinatorial stability, each node becomes aware of the complete network state with lower bandwidth. Since, cluster-head which have all the information it will sooner loss energy when information flow happens long time. So the energy of cluster-head is reduced. To overcome this problem HEFC algorithm is used which changes the cluster-head of the same cluster with higher energy. B. Higher Energy First clustering (HEFC) Algorithm

The following steps are carried out in HEFC algorithm:

HEF selects cluster heads according to the energy remaining for each node, and then the setup message (indicating cluster members, and the cluster head ID for each participated group) is sent to the cluster head of each cluster.

 The cluster head of each group broadcasts the setup message inviting the neighbour nodes to join its group. After receiving the setup message at this round, the regular node send the join message to its corresponding cluster head to commit to associate with the group. Each cluster head acknowledges the commitment, and send the schedule to its cluster members. Each node sends its energy information to its cluster head at the end of this clock cycle. From the performances of energy consumption models reported in the literature, it can be observed that the energy consumption for the cluster head is more than that for a regular node. it provides the ability to control the number of cluster heads used in each round, which makes the energy consumption analysis more accurate, and allows a tighter bound for the network lifetime to be found. In the proposed HEF scheme, a total of nodes with the highest residual energy will be chosen to be cluster heads in each round. The rest of the nodes would serve as regular nodes. Each regular node joins the cluster formed by the cluster head closest to it. The predefined energy level difference is used to enforce the cluster head rotation inside the cluster. Thus the energy for each cluster is maintained over time. EXPERIMENTAL RESULT Fig.2 represents the loss of packet due to link failure. When time increases packet drop also increases. When several nodes undergo transmission frequently, it may lead to loss of data or packet leakage during the transmission of data. By using Extended Fully Distributed Cluster Based routing, packet loss is greatly reduced for prolonged life time.

Fig2. Packet drop versus time

Fig.3 represents Cluster heads loses more energy as these are responsible for monitoring and updating the cluster QOS table as well as handling connection failures in the case of EFDCB algorithm. By using Higher Energy First Clustering algorithm, cluster heads in each cluster are changed after some time. Thus saving the energy of the network which in turn prolongs the network lifetime.

Fig3. Energy versus time

CONCLUSION The result is analyzed for both clustering algorithm. The performance of FDCB and EFDCB is reviewed. The Recovery time for EFDCB is more than two times faster than the global rerouting alternative at all failure rates tested. Thus main aim of Extended Fully Distributed Cluster-Based routing algorithm is to provide fault tolerance, which is a critical feature in providing QoS in the link failure-prone environment of mobile networks.Experiments show that a pure rerouting algorithm exhibits significant growth in recovery time as the source to destination distance increases. Though a fault tolerance property of QoS is achieved, the usage of energy by cluster head to transmit packet throughout the cluster may lower the energy of the cluster head. Cluster heads loses more energy as these are responsible for monitoring and updating the cluster QOS table as well as handling connection failures. By using Higher Energy First Clustering algorithm, cluster heads in each cluster are changed after some time. The predefined energy level difference is used to enforce the cluster head rotation inside the cluster. Thus the energy for each cluster is maintained over time.Thus saving the energy of the network. A smart cluster head selection strategy can significantly reduce energy consumption, which in turn prolongs the network lifetime.

REFERENCES [1] Larry C. Llewellyn, Kenneth M. Hopkinson and Scott R. Graham, Distributed Fault-Tolerant Quality of Wireless Networks, IEEE transactions on mobile computing, Vol. 10, No. 2, Feb 2011. [2] S. Chakrabarti and A. Mishra, QoS Issues in Ad Hoc Wireless Networks, IEEE Comm. Magazine, vol. 39, no. 2, pp. 142-148, Feb2001. [3] S. Chen and K. Nahrstedt, On Finding Multi-Constrained Paths,Proc. Record 1998 IEEE Intl Record on Comm. (ICC 98), pp. 874-879, 1998. [4] A.S. Nargunam and M.P. Sebastian, Fully Distributed Cluster Based Routing Architecture for Mobile Ad Hoc Networks, Proc. IEEE Intl Conf. Wireless and Mobile Computing, Networking, and Comm., pp. 383-389, 2005. [5] S. Nelakuditi, Z.L. Zhang, R.P. Tsang, and D.H.C. Du, Adaptive Proportional Routing: A Localized QoS Routing Approach, IEEE/ ACM Trans. Networking, vol. 10, no. 6, pp. 790-804, Dec. 2002. [6] A.H. Mohammad and M.E. Woodward, Localized Quality Based QoS Routing, Proc. Performance Evaluation of Computer and Telecomm. Systems (SPECTS), pp. 209-216, 2008. [7] A.S. Alzahrani and M.E. Woodward, End-to-End Delay in Localized QoS Routing, Proc. IEEE Intl Conf. Comm. Systems (ICCS), pp. 1700-1706, 2008. [8] P. Yang and B. Huang, QoS Routing Protocol Based on Link Stability with Dynamic Delay Prediction in MANET, Proc. Pacific- Asia Workshop Computational Intelligence and Industrial Applications (PACIIA), pp. 515-518, 2008. [9] A. Puri and S. Tripakis, Algorithms for Routing with Multiple Constraints, Report Number UCB/ERL M01/7, Electrical Eng. and Computer Science Dept., Univ. of California, 2001. [10] A.Mellouk, Quality of Service Dynamic Routing Schemes for Real Time Systems in IP Network, Proc. Networking Intl Conf. Systems and Intl Conf. Mobile Comm. and Learning Technologies ICN/ICONS/MCL, p. 93, 2006.

T.Saranya currently pursuing M.E Software Engineering in Bannari Amman Institute of Technology, Sathyamangalam, Tamilnadu, India. I am going to earn my Master degree at May 2012. I have completed B.E Computer Science in Veltech Engineering College, Chennai, Tamilnadu, India. I have earned my Bachelor degree on May 2010. I have presented papers for three International conferences all over Tamilnadu and I have published my paper in the top journals. My research interests lie in the area of wireless networks and mobile computing.