整机系统实时功率剖析与建模

发布时间：2018-03-30 06:25

本文选题：功率建模　切入点：功率剖析　出处：《计算机科学》2014年09期

【摘要】：功率剖析与建模是功率感知DBMS的基础。将主要硬件(处理器与磁盘)的利用率作为系统功率的指示器,根据资源利用率实时地估计系统功率,构建整机系统的实时功率模型。在多核架构下,整体CPU的活动信息会掩盖单个核的使用情况,因此,从执行核粒度考察执行频率与利用率的综合影响,采用多元线性回归方法拟合执行核的执行频率和利用率、磁盘的利用率和系统功率之间的关系。实验结果显示,所构模型平均相对误差小于12%,且占用系统资源较少,从而不会影响其他应用程序的执行,具有较好的应用价值。
[Abstract]:Power analysis and modeling is the basis of power sensing DBMS. The utilization ratio of main hardware (processor and disk) is used as the indicator of system power, and the system power is estimated in real time according to resource utilization. The real-time power model of the whole system is constructed. Under the multi-core architecture, the activity information of the whole CPU will cover up the use of the single core. Therefore, the comprehensive effect of the execution frequency and the utilization ratio is examined from the granularity of the execution kernel. The multivariate linear regression method is used to fit the relationship between the execution frequency and the utilization ratio of the core, the utilization ratio of the disk and the power of the system. The experimental results show that the average relative error of the model is less than 12, and the system resources are less occupied. So it will not affect the execution of other applications, and has good application value.
【作者单位】：浙江工业大学计算机科学与技术学院;浙江省可视媒体智能处理技术研究重点实验室;
【基金】：国家基金项目(61070042) 浙江省基金项目(LY13F020026,LY14F020017)资助
【分类号】：TP303;TP391.9

【相似文献】