An efficient adaptive strategy for searching in peer-to-peer networks

One of the main technical challenges in Peer-to-Peer (P2P) networks is how to efficiently locate desired resources. Although structured systems, based on distributed hash tables, can achieve fair effectiveness, they are not suitable for widely deployed Internet applications. In fact, this kind of systems shows many severe limitations, such as ignoring the autonomous nature of peers, and supporting only weakly semantic functions. Unstructured P2P networks are more attractive for real applications, since they can avoid both the limitations of centralized systems, and the drawbacks of structured approaches. However, their search algorithms are usually based on inefficient flooding schemes, that make large systems quickly overwhelmed by the query-induced load. In order to address this major limitation, this paper proposes a local adaptive routing protocol for searching in unstructured systems. The approach exploits a smart neighbor selection process that significantly improves resource discovery. Furthermore, this mechanism facilitates the dynamic evolution of a P2P system based on an unstructured topology, grouping together nodes with similar interests, thus allowing the emerging of small world topologies. Extensive simulations show that the algorithm proposed scales well and has a very good impact on the successful rate, allowing to retrieve the resources searched even when they are sparse.