一种基于统计分析的存储系统性能调优方法

Statistical Analysis-based Approach for Storage System Performance Tuning

计算机系统参数的合理配置能有效提升应用程序的性能。以NFS网络存储系统为例,提出了一种基于统计分析的存储系统性能调优方法,该方法分为关键系统参数识别和关键参数性能优化两个子阶段。阶段一采用方差分析(ANOVA)来建模系统参数的性能灵敏度,识别出对应用性能有显著影响的关键系统参数;然后在此基础上,阶段二采用响应面分析(RSM)来考察各关键参数对性能响应的影响,并综合前两个子阶段给出了性能调优算法,通过该算法找出系统的最优配置,从而最终达到性能调优的目的。最后,用实验评价了文中方法在Web,E-mail,Fileserver,Linux实用程序以及微基准测试等多种重要应用场景下的性能调优结果,实验结果证实了该调优方法的有效性和实用性。

The reasonable configuration of computer systems can dramatically improve performance of applications.Taking an NFS storage system as a case study,this paper proposed a statistical analysis-based performance tuning approach for storage systems and it proceeds in two phases.In the first phase,we leveraged the analysis-of-variance method（ANOVA）to model the performance sensitivity and thus identified the critical system parameters that have a significant effect on performance of applications;and furthermore,based on the previous phase,we employed the response surface methodology（RSM）to fulfill the tuning analysis of the critical parameters in the second phase.Combining the foregoing two steps,we then presented the performance tuning algorithm,which is eventually used to figure out the optimal combination of critical system parameters.Finally,the experimental results demonstrated the effectiveness and feasibility of our performance tuning approach through an extensive evaluation under various representative scenarios including Web,E-mail,Fileserver,Linux utilities,and micro-benchmarks.