基于回归模型的高端容错计算机TPC-C性能估算研究

摘要

高端容错计算机的TPC-C性能测试由于成本高昂且时间漫长,导致市场上只有少部分产品进行了该项测试,无法满足生产商和购买者的需求,高端容错计算机领域需要一种简便快捷、低成本的TPC-C性能估算方法.文中分析了影响TPC-C性能的各种因素,以近5年来发布了TPC-C测试结果的服务器为样本,利用数理统计的方法,在服务器TPC-C性能与硬件指标之间建立了线性回归模型.优化后的模型估算精度达到95％以上,在一定程度上解释了服务器的硬件指标与TPC-C性能之间的因果关系,具备了方便准确地估算TPC-C性能的现实意义.文中所提出的将数理统计的方法用于TPC-C性能估算的思路以及搜集的大量相关数据,对今后该项研究具有重要意义.%As the TPC-C performance test of high-end fault-tolerant computer cost expensively and long time,only a small part of the products on the market do the test,which can not meet the needs of producers and buyers.A simple,efficient and low-cost method for estimating the TPC-C performance is widely needed.This paper analyzes the various factor affecting the TPC-C performance,establishes a linear regression model between the TPC-C performance and hardware indicators using the method of mathematical statistics.The model takes the TPC-C test results released in the past five years as the sample.After tuning,the estimating accuracy of the model is more than 95％.The model explains the causal relationship between the server's hardware indicators and TPC-C performance to some extent,and estimates the TPC-C performance conveniently and accurately.The idea that using mathematical statistics for TPC-C performance estimation and the large volume of data collected has important significance for future study.