A Generalized Design for Database Heterogeneous Query Acceleration Based on FPGA

In recent years, the development of information technology has increased the performance demand of database. Improving the query efficiency of relational databases through FPGA heterogeneous design is becoming increasingly popular and has become an important direction for researchers in related fields to explore. This work deeply analyzes the principles and key steps of heterogeneous acceleration design for relational databases, and explores the compatibility and efficiency issues of existing FPGA heterogeneous query schemes in executing query applications. A generalized database heterogeneous query acceleration architecture based on FPGA is designed, and the database query operation mechanism of this architecture is introduced in detail. In order to further improve the efficiency of heterogeneous data query processing, we design a set of query acceleration kernel based on FPGA according to the practical application of database, which covers the common SQL query functions of relational database including sort, aggregation and join. In addition, the kernels involved in this work support variable width data queries, and multiple kernels adopt a pipeline design with high parallelism. The SQL query experiment on KaiwuDB platform, an open-source database, proves that the FPGA based heterogeneous database query acceleration architecture proposed in this work reduces the overall time overhead by up to 96.21 % compared to traditional software SQL query schemes.