Transparent Integration of a Dynamic FPGA Database Acceleration System

A substantial amount of recent work has been conducted on accelerating different database operators or database management systems (DBMS) as a whole, both in proprietary and open-source services. The missing piece of work in this field is to transparently accelerate a widely-used open-source database on an FPGA without substantial changes to the core database code. This PhD project aims to use a partially reconfigurable stream processing architecture to compose different database operators into a pipeline orchestrated automatically by the database query optimizer and query planner. This way, the database users do not have to change any of the existing database management system code while allowing a runtime system to optimize queries with information known only during runtime.