Communication cost minimization for SQL query mapping onto a shared-nothing multiprocessor architecture

This paper focusses on the minimization of intertask and inter-processor communication costs in the context of one SQL query mapping onto a shared-nothing architecture. After describing both the models of the application handled and the target multiprocessor architecture, a resolution method adapted to this particular problem is proposed: it is emphasized that only inter-processor communication costs can mainly be handled at mapping level, by attempting to minimize both inter-processor transfer times depending on the interconnection network topology, and the impact of pipeline starting/closing (start-up) times.