A performance study of declustering strategies for parallel spatial databases

Abstract

Effective exploitation of shared-nothing multiprocessors to improve the performance of spatial database applications requires novel query processing techniques. This is because, unlike traditional data, spatial objects are naturally multidimensional. As such, traditional single-dimensional declustering strategies are not directly applicable for spatial relation. In this paper, we propose and study the effect of several declustering strategies for spatial relations. We build a simulation model to compare the various strategies on the basis of throughput and response time for window queries. Our results show that one of the techniques which exploits redundancy performs best in most cases.