Research on distributed Hilbert R tree spatial index based on BIRCH clustering

Spatial index is the foundation of spatial database, while efficiency improvement of traditional serial spatial index has nearly reached its limit, it is therefore necessary to develop parallel spatial index approaches to break the bottleneck in accessing the root node in serial mode. This paper proposes a parallel spatial index called Hilbert R tree index, which can be carried on multicore CPU and computer cluster for parallel spatial queries and data retrieval. This new index method utilizes BIRCH clustering algorithm for spatial classification and data partition before distributed data deployment, considering geographic data characteristics; and creates spatial index in parallel environment by using of Hilbert filling curve. The test results demonstrate that parallel Hilbert R tree index based on BIRCH clustering algorithm can not only maintain internal spatial relations and the attributes of geographic dataset, but also has efficient performance in partitioning and retrieving spatial data.