负载感知的存储子系统调优研究  

Study on Workload-aware Tuning Methods of Storage Subsystem

作　　者：李益民[1] 邢春晓[1] 严琪[1] 胡庆成[1] 张小虎[1] 

机构地区：[1]清华大学计算机科学与技术系,北京100084

出　　处：《计算机科学》2007年第11期293-297,共5页Computer Science

基　　金：国家CNGI产业化类项目(分项CNGI-04-5-1D);中国博士后科学基金(20060390069);国家科技撑计划(2006BAH02A00)资助

摘　　要：存储IO的性能远远低于CPU、内存的性能,而且它们之间的差距还在扩大。因此,对于越来越多的数据密集型应用系统来说,存储系统往往是系统瓶颈。存储的性能不仅与存储子系统体系结构以及子系统中各部件的性能相关,还与系统的工作负载和应用环境相关。负栽感知的性能调优指存储子系统通过对负载特征的分析实现对应用环境的动态感知,并根据负栽特征动态调整系统运行策略。负载感知的性能调优使得存储子系统能够更合理地调度存储系统资源,从而提高IO性能。The I/O performance of storage sub-system is far lower than that of CPU or memory, and the gap between them is growing greater and greater. Storage subsystem is still the performance bottleneck of computer systems for data-intensive applications. The I/O performance of storage subsystem is associated with not only the architecture and the components of storage sub-system but workload characteristics and application circumstances as well. The core idea of workload sensitive performance-tuning is that storage subsystem continually analyzes workload characteristics so as to dynamically make sense of the application circumstances and automatically adiusts the systematical running strategy. Storage subsystem can reasonably schedule storage resources and considerably improve IO performance by Workload-Aware performance tuning.

关 键 词：存储子系统 负载感知 性能调优 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]