A Taxonomy-Based Approach for Constructing Semantics-Based Super-Peer Networks.

Clustering a peer-to-peer (P2P) network into distinct semantic clusters can efficiently improve the search efficiency and enhance scalability of the network. This paper considers P2P systems in which peers employ taxonomy hierarchy to describe the contents of their objects, and presents a taxonomy-based approach for constructing semantics-based super-peer networks. By dynamically clustering peers in taxonomy-based semantic space based on the semantics of their data and organizing the clusters into semantic routing overlays, an efficient query-routing algorithm can be used among these clusters. Preliminary evaluation indicates that our approach achieves a competitive trade-off between search latencies and overheads, and load-balancing among super-peers is well maintained. [ABSTRACT FROM AUTHOR]