Distributed caching and adaptive search in multilayer P2P networks

Abstract

To improve the scalability of Gnutella-like unstructured peer-to-peer (P2P) networks, a uniform index caching (UIC) mechanism was suggested in some earlier work. In UIC, query results are cached in all peers along the inverse query path such that the same query of other peers can be replied from their nearby-cached results. However, our experiments show that the UIC method causes a large amount of duplicated and unnecessary caching of items among neighboring peers. Aiming at improving the search efficiency, we propose a distributed caching mechanism, which distributes the cache results among neighboring peers. Furthermore, based on the distributed caching mechanism, an adaptive search approach is built which selectively forwards the query to the peers with a high probability of providing the desired cache results. All the enhancements above are defined in a protocol called distributed caching and adaptive search (DiCAS). In the DiCAS enhanced Gnutella network, all the peers are logically divided into multiple layers, with the character that all the peers in the same layer have the same group ID. The query flooding is restricted in one layer with the matched group ID. Our simulation study shows that, with the help of the index caching and search space division, the DiCAS protocol can significantly reduce the network search traffic in unstructured P2P systems without degrading query success rate.