Answering Constrained k-NN Queries in Unstructured P2P Systems

Abstract

The processing of k-NN queries has been studied extensively both in a centralized computing environment and in a structured P2P environment. However, the problem over an unstructured P2P system is not well studied despite of their popularity. Communication-efficient processing of k-NN queries in such an environment is a unique challenge due to the distribution, dynamics and large scale of the system. In this paper, we investigate the problem of efficiently computing k-NN queries over unstructured P2P systems. We first propose a location-based domination model to determine a search space. We then present two types of probing strategies, radius-convergence and radius-expanding. A comprehensive performance study demonstrates that our techniques are efficient and scalable.