Metric-Based Similarity Search in Unstructured Peer-to-Peer Systems

Peer-to-peer systems constitute a promising solution for deploying novel applications, such as distributed image retrieval. Efficient search over widely distributed multimedia content requires techniques for distributed retrieval based on generic metric distance functions. In this paper, we propose a framework for distributed metric-based similarity search, where each participating peer stores its own data autonomously. In order to establish a scalable and efficient search mechanism, we adopt a super-peer architecture, where super-peers are responsible for query routing. We propose the construction of metric routing indices suitable for distributed similarity search in metric spaces. Furthermore, we present a query routing algorithm that exploits pruning techniques to selectively direct queries to super-peers and peers with relevant data. We study the performance of the proposed framework using both synthetic and real data demonstrate its scalability over a wide range of experimental setups.