Evaluation of Disk Allocation Methods for Parallelizing Spatial Queries on Grid Files‡

Spatial Database Systems are characterized by large amounts of geometric and geographic data. Query response times in these systems are crucial, since these systems are often used interactively for decision support systems. The Grid file[1] is a well-known spatial access method that has great potential for parallelism, which reduces the response time of spatial queries for time-critical on-line applications of spatial databases. An important problem in parallelizing a grid file is the choice of a disk allocation method for mapping the data-blocks to multiple disks in the parallel secondary storage. This paper explores the disk allocation methods used to allocate the data pages of a grid file among a set of disks that can be accessed in parallel. In this environment, given N disks, a perfect allocation will speed up the processing of each query by a factor of N. The paper shows that no disk allocation is perfect for the set of all orthogonal range queries, even on uniformly distributed read-only data. We then introduce two families of allocation methods, namely the Linear allocation method and the Lattice allocation method, which are perfect for a large collection of interesting path queries (rows and columns) and range queries (small rectangles), on an interesting set of data distributions. We address the issues in extending disk allocation methods to general data distributions that are updated randomly. Finally, we provide experimental results on both the performance of the proposed methods and other well known disk allocation methods on different query sets, data distributions and data set sizes.