Parallel bulk-loading of spatial data

摘要

Spatial database systems have been introduced in order to support non-traditional data types and more complex queries. Although bulk-loading techniques for access methods have been studied in the spatial database literature, parallel bulk-loading has not been addressed in a parallel spatial database context. Therefore, we study the problem of parallel bulk-loading, assuming that an R-tree like access method need to be constructed, from a spatial relation that is distributed to a number of processors. Analytical cost models and experimental evaluation based on real-life and synthetic datasets demonstrate that the index construction time can be reduced considerably by exploiting parallelism. I/O costs, CPU time and communication costs are taken into consideration in order to investigate the efficiency of the proposed algorithm.