Query Based Clustering Method in Structured P2P Overlay Networks

In structured peer-to-peer (P2P) overlay networks, similar documents are randomly distributed over peers with their data identifiers consistently hashed, which makes complex search challenging. Current state-of-the-art complex query approaches in structured P2P systems are mainly based on inverted list intersection. When the identifiers are distributed among peers, a complex query may involve many peers and cause a large amount of network traffic. One solution of implementing efficient complex query is to organize documents on each peer using clustering. In this paper, we propose a clustering method, QBC, which is composed of pull mode and push mode. Pull mode uses historical queries to direct clustering in structured P2P overlay networks and push mode applies modified vector space model (VSM) to define document set on each peer in order to assist clustering. Experiments show that QBC can reduce the number of peers visited during complex search, hence both query response time and network traffic are decreased.