GPU Acceleration for SQL Queries on Large-Scale Distributed Systems

Abstract

General purpose GPUs are a powerful hardware with a number of applications in the realm of relational databases. We further extended a database framework designed to allow for GPU execution queries. Our technique is novel in that it implements Dynamic Parallelism, a new feature in recent hardware, to accelerate SQL JOINs. Query execution results in 1.25X speedup on average with respect to a previous method, also accelerated by GPUs, which employs a multi-dimensional CUDA Grid.More importantly, we divided the queries to run on multiple BW nodes to investigate the scalability of both SELECT and JOIN.