Adaptive peer-to-peer routing with proximity

Abstract

In this paper, we presented a routing strategy for requests in unstructured peer-to-peer networks. The strategy is based on the adaptive routing Q-routing. The strategy uses reinforcement learning to estimate the cost of routing a request. Such a strategy is scalable only if the routing indices are of reasonable size. We proposed and comparatively evaluated three methods for the pruning for the pruning of the routing indices. Our experiments confirm the validity of the adaptive routing and the scalability of a pruning approach based on a pruning strategy considering the popularity of the resources. © Springer-Verlag Berlin Heidelberg 2003.