Towards self-tuning data placement in parallel database systems

Parallel database systems are increasingly being deployed to support the pe rformance demands of end-users. While declustering data across multiple nod es facilitates parallelism, initial data placement may not be optimal due t o skewed workloads and changing access patterns. To prevent performance deg radation, the placement of data must be reorganized, and this must be done on-line to minimize disruption to the system. In this paper, we consider a dynamic self-tuning approach to reorganization in a shared nothing system. We introduce a new index-based method that fac iliates fast and efficient migration of data. Our solution incorporates a g lobally height-balanced structure and load tracking at different levels of granularity. We conducted an extensive performance study, and implemented t he methods on the Fujitsu AP3000 machine. Both the simulation and empirical results demonstrate that our proposed method is indeed scalable and effect ive in correcting any deterioration in system throughput.