SPUN: A P2P Probabilistic Search Algorithm Based on Successful Paths in Unstructured Networks

Efficient searching for information is an important goal in peer-to-peer (P2P) networks. Searching in an unstructured P2P network is particularly challenging due to the random nature of the P2P overlay links. In this paper, we propose a novel probabilistic search mechanism called SPUN, that increases the success ratio of queries while keeping the bandwidth consumption considerably low. SPUN is an informed search mechanism that improves upon state-of-art probabilistic mechanism, namely, the Adaptive Probabilistic Search (APS). The core principle of our algorithm is to exploit the successful query paths that develop during the lifetime of P2P network and converge toward the target objects. Our work introduces a new neighbor selection criterion which allows a peer to evaluate its neighbors based on the strength of successful paths the neighbor leads to. We also introduce a peer profile exchange mechanism that supports the reduction in uncertainty in peer selection decision. Our extensive simulation results confirm that our path-based algorithm performs 25% better than APS and several of its variants and is capable of achieving higher success ratios with fewer walkers each with an average message size of 71 bytes.