基于性能计数事件的计算机功耗估算模型

摘要

Accurately to get computer power consumption is the key of energy optimization,thus this paper proposed a new method to estimate computer power consumption.Based on hardware performance events and machine leaning theory,this paper built a power estimation model to estimate the computer node power consumption.When building the power estimation model,it used multiple linear regression and support vector regression methods and made a comparison.Research result indicates that power model based on hardware performance events can accurately estimate real time power with an error less than 3％.Compared with other similar existed models,the model is more accurate and more general.%精准快速获取计算机系统的实时功耗是功耗优化研究的基础,因此提出并建立了一种高精度的计算机功耗估算模型.通过分析统计系统运行时代表性的性能计数事件,应用机器学习理论分析性能事件与功耗的关系,建立多核计算机系统实时功耗估算模型.模型构建时使用多元线性回归(multiple linear regression,MLR)方法以及支持向量回归(support vector regression,SVR)方法分析两者关系,并对两种方法建立的功耗估算模型进行了对比分析.实验结果表明,基于性能事件的功耗估算模型可准确估计计算机实时功耗,估算误差不高于3％.与已有模型相比较,该估算模型精度更高、通用性更好.