计算机科学 ›› 2014, Vol. 41 ›› Issue (9): 32-37.doi: 10.11896/j.issn.1002-137X.2014.09.005

• 综述 • 上一篇    下一篇

整机系统实时功率剖析与建模

杨良怀,朱红燕   

  1. 浙江工业大学计算机科学与技术学院 杭州310023;浙江省可视媒体智能处理技术研究重点实验室 杭州310023;浙江工业大学计算机科学与技术学院 杭州310023
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受国家基金项目(61070042),浙江省基金项目(LY13F020026,LY14F020017)资助

Whole System Realtime Power Profiling and Modeling

YANG Liang-huai and ZHU Hong-yan   

  • Online:2018-11-14 Published:2018-11-14

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。

关键词: 功率建模,功率剖析,软功率计,多元回归

Abstract: Power profiling and modeling are fundament to power aware DBMS.This paper proposed a component-level power modeling method by exploiting the utilizations of the main components (CPU and disk) as the indicators of the system power.To cope with the multi-core architecture,we investigated into the individual cores instead of the monolithic CPU that may cover up some essential details.And we fitted the relationship between the utilizations and power by using the multiple linear regression method.Experimental results show that the average relative error of the power model is less than 12%,and it is lightweight without incurring too much additional cost on other applications.

Key words: Power modeling,Power profiling,Soft-power-meter,Multiple regression

[1] Wu C F.Making a case for efficient supercomputing [J].Queue,2003,1(7):54-64
[2] Heng Z,Ellis C S,Lebeck A R,et al.ECOSystem:managing e-nergy as a first class operating system resource[J].SIGARCH Comput.Archi.News,2002,30(5):123-132
[3] Agrawal R,Ailamaki A,Bernstein P A,et al.The Claremont report on database research[J].SIGMOD Record,2008,37(3):9-19
[4] Flinn J,Satyanarayanan M.Energy-aware adaptation for mobileapplications[C]∥Proceedings of the ACM symposium on Opera-ting systems principles.1999:48-63
[5] Bellosa F.The benefits of event-driven energy accounting inpower-sensitive systems[C]∥Proceedings of the Workshop on ACM SIGOPS.2000:37-42
[6] Li T,John L K.Run-time modeling and estimation of operating system power consumption[C]∥Proceedings of the ACM International Conference on Measurement and Modeling of Computer Systems.2003:160-171
[7] Bircher W L,John LK.Complete System Power Estimation:A Trickle-Down Approach Based on Performance Events[C]∥Performance Analysis of Systems & Software.2007:158-168
[8] Economou D,Rivoire S,Kozyrakis C.Full-system power analysis and modeling for server environments[C]∥Workshop on Modeling,Benchmarking and Simulation.2006
[9] Do T,Rawshdeh S,Shi W S.pTop:A process-level power profiling tool[C]∥Proceedings of the Workshop on Power Aware Computing and Systems.2009
[10] Chen H,Li Y,Shi W S.Fine-grained power management using process-level profiling[J].Sustainable Computing:Informatics and Systems,2012,2(1):33-42
[11] 李晓,李战怀,刘文洁,等.计算机系统实时估算方法:中国,CN102221874 A[P].http://www.google.com/patents/CN102221874A?cl=zh,2011-10-19
[12] Laudon J.Performance/Watt:The New Server Focus[J].SIGARCH Computer Architecture News,2005,33(4):5-13
[13] Geer D.Chip makers turn to multicore processors[J].Computer & Processing,2005,38(5):11-13
[14] Bostoen T,Mullender S,Berbers Y.Power-Reduction Tech-niques for Data-Center Storage Systems[J].ACM Comput.Surv.,2013,45(3):1-38
[15] Clam AntiVirus.http://www.clamav.net/

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!