Searching with Feature Similarity in Unstructured Peer-to-Peer Networks

Abstract

This paper proposes a semantic overlay architecture design in unstructured peer-to-peer network environments. According to semantic distances based on feature similarity of files in a system, this design organizes a semantic peer-to-peer overlay that is capable of delivering query messages to neighboring peers in semantic proximity. In addition, this paper develops a deterministic search approach that can perform content search in a parallel manner over a semantic overlay. Extensive simulations are conducted to examine the proposed deterministic search mechanism in unstructured peer-to-peer networks. Experimental results show that this mechanism can provide satisfactory search services with cost-effective performance in terms of message cost, search latency and hit ratio.