Speeding up Spatial Database Query Execution using GPUs

• DOI: 10.1016/j.procs.2012.04.205
• Publication Date: 2012-06-02T16:19:19Z
• Authors: Bogdan Simion; Suprio Ray; Angela Demke Brown
• Keywords: Speedup; Spatial Query; Graphics processing unit

Spatial databases are used in a wide variety of real-world applications, such as land surveying, urban planning, and environmental assessments, well geospatial Web services. As uses spatial become more widespread, there is growing need for good performance applications. In workloads, queries tend to be computationally-intensive due the complex processing geometric relationships. Furthermore, significant fraction query execution time spent on CPU stalls memory accesses, caused by ever-increasing processor-memory speed gap. With advent massively-parallel graphics-processing hardware (GPUs) frameworks like CUDA, opportunities speeding up have emerged. addition massive parallelism, GPUs can also better hide latency.We aim using CUDA recent GPU cards. One main challenges transfer from memory. We implement set six typical achieve baseline speedup (without cost) 62-318x over counterparts. show that cost amortized each individual query. For simpler queries, time, but we still 6-10x speedup. becomes negligible compared obtain 62-240x