No abstract

Placing the processing power near the data, rather than shipping the data to the processor is inevitable and demanding in the era of Big Data. With a given near-data processor, it is important to use it properly to reduce the massive data transfer between data sources and computing nodes. In database query processing, as many operations as possible should be sent to the near-data processor, because it can process them at line-rate and thus can eliminate massive data transfer across the network. Unfortunately, there is still not enough support for effective utilization of hardware capabilities to solve the issues of data processing in query processors. In this paper, we propose a new approach to the query optimization in a relational DBMS, which considers the specialized computing capabilities of the attached FPGA-based near-data processor. We use extended rules to achieve hardware-conscious optimization and refine the optimization strategies according to changes in the hardware state. Our evaluations demonstrate that the proposed query-optimization approach can improvethe processing of queries.