计算机CPU服务能耗指标的相关性分析  

The Correlation Analysis for the Service Energy Consumption Index of CPU

作　　者：翟畅[1,2] 何芳[3,4] 赵彤洲[1,2] 周萍[1,2] 李慕[1,2] Zhai Chang;He Fang;Zhao Tongzhou;Zhou Ping;Li Mu(School of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,Hubei;Hubei Key Laboratory of Intelligent Robot,Wuhan 430205,Hubei;Hubei University of Technology,Wuhan 430640,Hubei;Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences,Wuhan 430071,Hubei)

机构地区：[1]武汉工程大学计算机科学与工程学院,湖北武汉430205 [2]智能机器人湖北省重点实验室,湖北武汉430205 [3]湖北工业大学计算机学院,湖北武汉430064 [4]中国科学院武汉物理与数学研究所,湖北武汉430071

出　　处：《电脑与电信》2016年第12期8-11,共4页Computer & Telecommunication

基　　金：国家自然科学基金资助项目;项目编号:61103136

摘　　要：由计算机的广泛应用带来的大量计算任务将导致能量消耗增高。计算机CPU承担了主要计算任务,因此,分析CPU的服务能耗指标,准确找到影响能耗的重要因素能为面向绿色效能的服务选择提供依据。本文利用能耗监测仪采集到的数据,构建了多元回归模型及单变量和多变量相关性分析,并通过拟合优度进行评估,实验表明,在利用马氏距离剔除利群点后构建的多元回归模型能很好地拟合整体样本数据。The widespread use of computers brings a large number of computational tasks, leading to the increasing energy consumption.CPU undertakes the main task of computing, so it is necessary to analyze the energy consumption index to search the important factors that have influence to energy consumption. It can provide the basis for the service selection of green efficiency. This paper uses the energy consumption monitor to collect the data, constructs multiple regression models and makes univariate and multivariate correlation analysis which uses the goodness of fit to evaluate the models. Experiments show that the multivariate model can well fit the whole sample data after excluding the outliers with Mahalanobis distance.

关 键 词：离群点 马氏距离 多元回归 拟合优度 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]