A Stochastic Clustering Algorithm for Swarm Compact Routing

Abstract

The Internet community is still looking for a long term solution for their interdomain routing scalability issues. Compact routing based Algorithms are promising candidates. Cowen's compact routing scheme is the first stretch-3 scheme found. This paper presents a new swarm-based stochastic clustering algorithm based on Cowen's scheme. By simulation and analysis, the stretch distribution of all paths generated by the clustering algorithm is derived. The algorithm is applied as an compact routing scheme on an Internet-like inter-domain topology with 3000 autonomous systems. Conservative analysis (ignoring potential "short cut" paths) output an average stretch of 1.04 and an maximum stretch of 3. Further investigations show that more than 90% of all paths have stretch-1. By manual check of all stretch-3 paths, it is found that the source nodes of these paths have more immediate neighbors than they may include in their neighborhood due to Cowen's neighborhood size limit. The simulation results show that each node has around 240 routing entries in total, which is indeed comparable to the Cowen's original scheme.