DPR-tree: a distributed parallel spatial index structure for high performance spatial databases
Yan Jie Zhou, Qing Xin Zhu, Qiang Liu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
Parallelism of spatial index could significantly improve the performance of spatial queries, special for massive spatial databases, so the research of parallel spatial index takes a important role in high performance spatial databases. Existing parallel spatial index methods have two main shortcoming: one is accessing hotspot and bottleneck of index items located in main server, the other is high costs and complicated operations for maintaining index consistency. Aim at these, a distributed parallel spatial index structure called DPR-tree is proposed. It splits whole index region into partition sub-regions by using Hilbert space-filling curve grid and organizes index sub-regions according to locality of spatial objects, then maps index sub-regions to partition sub-regions and assigns these index sub-regions to different computer nodes by a appointed map function, Each computer node manages a multi-level distributed sub-Rtree which is built from a index sub-region. Our experimental results indicate that the proposed parallel spatial index can achieve speedup well and offer significant potential for reducing query response time.