A Robust and Self-tuning Page-Replacement Strategy for Spatial Database Systems

For a spatial database management system, it is an important goal to minimize the I/O-cost of queries and other operations. Several page-replacement strategies have been proposed and compared for standard database systems. In the context of spatial database systems, however, the impact of buffing techniques has not been considered in detail, yet. In this paper, different page-replacement algorithms are compared for performing spatial queries. This study includes well-known techniques like LRU and LRU-K as well as new algorithms observing spatial optimization criteria. Experiments show that spatial pagereplacement algorithms outperform LRU buffers for many distributions, but not for all investigated query sets. Therefore, a combination of spatial page-replacement strategies with LRU strategies is proposed and experimentally investigated. An algorithm is presented, which is self-tuning and adapts itself to different or changing query distributions. This adaptable spatial buffer outperforms LRU in respect to the I/O-cost by performance gains of up to 25%.