A Location-Aware Efficient Content-Based Searching over Unstructured P2P Network

Abstract

With the explosion of P2P (peer-to-peer) applications, searching and locating necessary content is a fundamental part for utilizing Internet information and resources. Unstructured P2P network suffers from the topology mismatch problem between the overlay networks and the underlying physical network and suffers from low searching efficiency based mostly on keys. We suggest an approach of combining location-aware topology matching and interest-based searching to solve the mismatching topology problem and to improve searching efficiency. Key techniques in our approach are constructing location-aware topology based on building an overlay minimum spanning tree (MST) among each source peer and the peers within specific hops, and further optimizing the neighbor connections outside the tree, and building location and interest-based content correlation subnets and routing strategy using combination of peers' locations and metadata-based content correlation expression model and active learning algorithm. The results show that our method achieves approximately 29% reduction on traffic cost and about 60% reduction on query response time.