Evolutionary Local Search for Designing Peer-to-Peer Overlay Topologies Based on Minimum Routing Cost Spanning Trees

Finding overlay topologies for peer-to-peer networks on top of the Internet can be regarded as a network design problem, in which a graph with minimum communication costs is desired. An example of such a graph is a spanning tree connecting all nodes in the overlay. We present evolutionary algorithms incorporating local search for the minimum routing cost spanning tree problem in which the overall routing/communication cost is minimized. We present three types of local search for this problem as well as an evolutionary framework for finding (near)optimal solutions to the problem. Moreover, we present results from a fitness landscape analysis for the three types of local optima that reveal interesting properties of the problem data based on real measurements in the Internet. We demonstrate that our proposed algorithms find near optimum solutions reliably by comparing against a lower bound of