Peer-to-peer unstructured anycasting using correlated swarms

Over the recent years, social network analysis has received renewed interest because of the significant increase in the number of users relying on applications based on them. An important criterion for the success of any social-networking based application is the efficiency of search. In this paper, we propose and analyze a method of anycast search based on correlated communities or subgroups, i.e., using group-to-group caching. It works by restricting search to peers that belong to communities which are highly correlated with the requested community. We analytically prove that our proposed method works better than basic random walk, which remains a widely used method for performing search in these networks. Indeed our experiments prove that the proposed method reduces the search time by as much as 30% to that based on random walk. Our experiments also indicate that the proposed method outperforms basic random walk even under considerable peer-churn.