A Cluster Procedure for the Dynamic Source Routing Protocol in Ad hoc Networks

Numerous studies have shown the difficulty for a routing protocol to scale to large wireless networks. Ad hoc routing protocols have to cope with an increasing node density and a significant number of nodes. DSR has been supported by the Manet working group as an IP routing protocol adapted to multi-hop wireless ad hoc networks [1,2]. DSR is a reactive source routing protocol. Source node provides the routing information to reach the destination in the header of each data packet. Routes are computed on demand by flooding the network with request packets. DSR is one of the most efficient ad hoc protocol regarding to routing overhead minimisation [3]. However, as stated before, the flooding of the whole network with routing packets generates significant overhead. Therefore, numerous hierarchical routing schemes have been designed to make ad hoc protocols more scalable. Clustering algorithms are based on the partition of the network into groups. Then, a cluster head is selected in each group. To enhance the scalability of the DSR, we propose in this paper an extension of its Route Discovery mechanism named Cluster Source Routing (CSR) rather than a new protocol. It is an hybrid of a reactive and a proactive method which relies on clustering procedures. We assume this option is totally transparent with regard to DSR to ensure full compatibility between native DSR and DSR-CSR nodes. The DSR packet format is preserved and clustering procedures are completed through the DSR option mechanisms. CSR is designed to be adaptive according to network dynamics (its use can depend on the network size or the mobility degree): the switch between the DSR and DSR-CSR modes can be carried out if required. The main idea of our proposition is to transfer the Route Discovery procedure to the 1-server level to prevent the network flooding due to the DSR Route Discovery. The CSR is based on a 2-level hierarchical scheme. The first level of cluster is the 0-cell cluster: each node of the cell is 1-hop away from the Cluster Head. The 1-server cluster gathers a set of cells which leader is the Server. A combination between a mobility metric and a connectivity metric is our criterion to select the clusters leaders. The Server periodically questions the CHs to update the routing information and maintains the 1-server cluster. CSR uses the hierarchy to perform Route Discovery. Thus, Route Discovery does not require flooding mechanism and overhead is minimised. Our evaluation study is mainly focused on the overhead measurement and the comparison of the native DSR Route Discovery with the DSR-CSR one. Parameters of simulation are mobility degree, node density and network connectivity degree. Simulations show the benefit of the CSR in dense network configurations (as the node density and the connectivity degree increase). This poster proposes the CSR extension to improve the DSR scalability. Besides, the use of an upper level of route cache allows route computation with QoS constraints. Simulations on NS2 are under way to evaluate thoroughly the overhead introduced by clustering mechanisms [4].