Metric Overlay Networks: Processing Similarity Queries in P2P Databases

In this paper we present some preliminary results on the processing of similarity queries in a self-organized Metric Overlay Network (MON). A MON is an unstructured peer-to-peer network where peers choose their neighbors according to a notion of similarity which is defined through a metric distance. We compare four different routing techniques and investigate how information about peers and queries can be exploited to increase the quality of the results. In particular, our experiments show that major improvements can be observed over naive query processing strategies: such promising results make MONs attractive for searching in widely distributed databases of multimedia objects.