SwiftSpatial: Spatial Joins on Modern Hardware

Spatial joins are among the most time-consuming queries in spatial data management systems. In this paper, we propose SwiftSpatial, a specialized accelerator architecture tailored for spatial joins. SwiftSpatial contains multiple high-performance join units with innovative hybrid parallelism, several efficient memory management units, and an integrated on-chip join scheduler. We prototype SwiftSpatial on an FPGA and incorporate the R-tree synchronous traversal algorithm as the control flow. Benchmarked against various CPU and GPU-based spatial data processing systems, SwiftSpatial demonstrates a latency reduction of up to 5.36x relative to the best-performing baseline, while requiring 6.16x less power. The remarkable performance and energy efficiency of SwiftSpatial lay a solid foundation for its future integration into spatial data management systems, both in data centers and at the edge.