Performance of database workloads on shared-memory systems with out-of-order processors

摘要: Database applications such as online transaction processing (OLTP) and decision support systems (DSS) constitute the largest fastest-growing segment of market for multiprocessor servers. However, most current system designs have been optimized to perform well on scientific engineering workloads. Given radically different behavior database workloads (especially OLTP), it is important re-evaluate key design decisions in context this class applications.This paper examines shared-memory multiprocessors with aggressive out-of-order processors, considers simple optimizations that can provide further performance improvements. Our study based detailed simulations Oracle commercial engine. The results show combination execution multiple instruction issue indeed effective improving workloads, providing gains 1.5 2.6 times over an in-order single-issue processor OLTP DSS, respectively. In addition, speculative techniques enable implementations memory consistency models significantly improve stricter models, bringing within 10--15% more relaxed models.The second part our focuses challenging workload. We stream buffer reducing remaining stalls OLTP, a 17% reduction time (approaching perfect cache 15%). Furthermore, characterization shows large fraction data communication misses exhibit migratory behavior; preliminary software prefetch writeback/flush hints be used reduce by 12%.