学术文献库

Hardware acceleration of database query processing can be done with the help of FPGAs. In particular, they are partially reconfigurable at runtime, which allows for the runtime adaption of the hardware to a variety of queries. Reconfiguration itself, however, takes some time. As the affected area of the FPGA is not available for computations during the reconfiguration, avoiding some of the reconfigurations can improve overall performance. This paper presents optimizations based on query sequences, which reduces the impact of the reconfigurations. Knowledge of upcoming queries is used to (I) speculatively start reconfiguration already when a query is still running and (II) avoid overwriting of reconfigurable regions that will be used again in subsequent queries. We evaluate our optimizations with a calibrated model and measurements for various parameter values. Improvements in execution time of up to 28% can be obtained even with sequences of only two queries.