Research On Data Link And Protocol Technology For Database Acceleration

As the amount of data on the database continues to increase,the information processing system has more and more demand for rapid query of massive data,and its performance requirements are also getting higher and higher.In addition to multi-core CPU for database acceleration,heterogeneous systems with FPGA,GPU,ASIC and other chips can make full use of their parallel computing capabilities and improve data query efficiency,which has become a development trend of database acceleration.This paper focuses on the heterogeneous database acceleration system composed of CPU and FPGA.Because the data is stored in a discontinuous physical address of the CPU,the data link between its device driver software and the FPGA hardware query algorithm and its multi-queue protocol are designed.According to the protocol,the hardware logic architecture is designed and each queue protocol resolution module is implemented,and the correctness of the function is verified by the behavior simulation.The use of dynamic partial reconfiguration to achieve the expansion of database acceleration functions such as sorting,connection,and filtering.It is verified by experiments that in the multi-thread concurrent operation read-write queue,the highest read or write bandwidth is over 3150MBps,which is more than 84.5%of the theoretical speed.And the speed meets the design requirements.When a large block is transmitted,the data link exhibits high performance,but when the data block is small,performance is degraded.