Accelerating data filtering for database using FPGA

Abstract

Abstract In the big data era, in order to relieve computational pressure on overloaded CPU caused by ever increasing amount of data, many researches focus on hardware acceleration using FPGA for data-intensive applications. In this paper, a novel FPGA-based storage engine is proposed for DBMS in the cloud with focus on data filtering operation. A hardware data filter is designed which can significantly speedup filtering operations by utilizing parallelism provided by FPGA. Meanwhile, it can support different queries without partial reconfiguration. This FPGA-based storage engine is integrated with DBMS to realize end-to-end acceleration. In addition, an intelligent filtering on/off switch is designed to adaptively decide whether the FPGA-based filter should be employed, based on selectivity estimation. Experimental results show that the proposed solution realizes on average 2.80x computation speedup for data filtering compared with the software baseline, and achieves up to 1.95x improvement in end-to-end evaluation compared with conventional storage engine in low-selectivity cases. Moreover, the FPGA-based solution achieves 2.87x improvement on energy efficiency compared with the similar GPU-based acceleration solution.