Optimization Strategy of Parallel Query Processing Based on Multi-core Architecture.

Yongheng Chen,Wanli Zuo,Fengling He,Kerui Chen

DOI: https://doi.org/10.4156/jcit.vol5.issue8.24

2010-01-01

Abstract:Chip Multi-Processor (CMP) could support more than two threads to execute simultaneously, and some executing units are owned by each core. Because threads share various resources of CMP, such as L2-Cache, among many threads, CMP system is inherently different from multiprocessors system and, CMP is also different from simultaneously multithreading (SMT). In this paper a novel and complete approach on how to parallelism for relational database multithreaded query execution that strives for maximum resource utilization for both CPU and disk activities. The focus of this approach is on how to use the multithreaded parallel technique to optimize and process queries based on multicore architecture. A set of algorithms for implementing and optimizing the best query plan, such as the algorithms for scheduling and parallel executing the query plan, the algorithms for allocating thread to sub-query and memory to the buffers between operations in pipelining execution, are proposed in this method. Additionally this paper analysis and optimize the implementation of parallel buffers and multithread. In the experiments, this paper evaluates performance of the parallel buffers and tests the coordination of the multithread.