How much power does your server consume? Estimating wall socket power using RAPL measurements

Kashifnizam Khan (Corresponding Author), Zhonghong Ou, Mikael Hirki, Jukka K. Nurminen, Tapio Niemi

Research output: Contribution to journalArticleScientificpeer-review

5 Citations (Scopus)

Abstract

Full system electricity intake from the wall socket is important for understanding and budgeting the power consumption of large scale data centers. Measuring full system power, however, requires extra instrumentation with external physical devices, which is not only cumbersome, but also expensive and time consuming. To tackle this problem, in this paper, we propose to model wall socket power from processor package power obtained from the running average power limit (RAPL) interface, which is available on the latest Intel processors. Our experimental results demonstrate a strong correlation between RAPL package power and wall socket power consumption. Based on the observations, we propose an empirical power model to predict the full system power. We verify the model using multiple synthetic benchmarks (Stress-ng, STREAM), high energy physics benchmark (ParFullCMS), and non-trivial application benchmarks (Parsec). Experimental results show that the prediction model achieves good accuracy, which is maximum 5.6Â % error rate.
Original languageEnglish
Pages (from-to)207-214
Number of pages8
JournalComputer Science: Research and Development
Volume31
Issue number4
DOIs
Publication statusPublished - 1 Nov 2016
MoE publication typeA1 Journal article-refereed

Fingerprint

Servers
Electric power utilization
Budget control
High energy physics
Electricity

Keywords

  • energy efficiency
  • HPC
  • power modeling
  • RAPL

Cite this

Khan, Kashifnizam ; Ou, Zhonghong ; Hirki, Mikael ; Nurminen, Jukka K. ; Niemi, Tapio. / How much power does your server consume? Estimating wall socket power using RAPL measurements. In: Computer Science: Research and Development. 2016 ; Vol. 31, No. 4. pp. 207-214.
@article{75218d1f9f794a258db944a469b180c9,
title = "How much power does your server consume? Estimating wall socket power using RAPL measurements",
abstract = "Full system electricity intake from the wall socket is important for understanding and budgeting the power consumption of large scale data centers. Measuring full system power, however, requires extra instrumentation with external physical devices, which is not only cumbersome, but also expensive and time consuming. To tackle this problem, in this paper, we propose to model wall socket power from processor package power obtained from the running average power limit (RAPL) interface, which is available on the latest Intel processors. Our experimental results demonstrate a strong correlation between RAPL package power and wall socket power consumption. Based on the observations, we propose an empirical power model to predict the full system power. We verify the model using multiple synthetic benchmarks (Stress-ng, STREAM), high energy physics benchmark (ParFullCMS), and non-trivial application benchmarks (Parsec). Experimental results show that the prediction model achieves good accuracy, which is maximum 5.6{\^A} {\%} error rate.",
keywords = "energy efficiency, HPC, power modeling, RAPL",
author = "Kashifnizam Khan and Zhonghong Ou and Mikael Hirki and Nurminen, {Jukka K.} and Tapio Niemi",
year = "2016",
month = "11",
day = "1",
doi = "10.1007/s00450-016-0325-4",
language = "English",
volume = "31",
pages = "207--214",
journal = "SICS Software-Intensive Cyber-Physical Systems",
issn = "2524-8529",
publisher = "Springer",
number = "4",

}

How much power does your server consume? Estimating wall socket power using RAPL measurements. / Khan, Kashifnizam (Corresponding Author); Ou, Zhonghong; Hirki, Mikael; Nurminen, Jukka K.; Niemi, Tapio.

In: Computer Science: Research and Development, Vol. 31, No. 4, 01.11.2016, p. 207-214.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - How much power does your server consume? Estimating wall socket power using RAPL measurements

AU - Khan, Kashifnizam

AU - Ou, Zhonghong

AU - Hirki, Mikael

AU - Nurminen, Jukka K.

AU - Niemi, Tapio

PY - 2016/11/1

Y1 - 2016/11/1

N2 - Full system electricity intake from the wall socket is important for understanding and budgeting the power consumption of large scale data centers. Measuring full system power, however, requires extra instrumentation with external physical devices, which is not only cumbersome, but also expensive and time consuming. To tackle this problem, in this paper, we propose to model wall socket power from processor package power obtained from the running average power limit (RAPL) interface, which is available on the latest Intel processors. Our experimental results demonstrate a strong correlation between RAPL package power and wall socket power consumption. Based on the observations, we propose an empirical power model to predict the full system power. We verify the model using multiple synthetic benchmarks (Stress-ng, STREAM), high energy physics benchmark (ParFullCMS), and non-trivial application benchmarks (Parsec). Experimental results show that the prediction model achieves good accuracy, which is maximum 5.6Â % error rate.

AB - Full system electricity intake from the wall socket is important for understanding and budgeting the power consumption of large scale data centers. Measuring full system power, however, requires extra instrumentation with external physical devices, which is not only cumbersome, but also expensive and time consuming. To tackle this problem, in this paper, we propose to model wall socket power from processor package power obtained from the running average power limit (RAPL) interface, which is available on the latest Intel processors. Our experimental results demonstrate a strong correlation between RAPL package power and wall socket power consumption. Based on the observations, we propose an empirical power model to predict the full system power. We verify the model using multiple synthetic benchmarks (Stress-ng, STREAM), high energy physics benchmark (ParFullCMS), and non-trivial application benchmarks (Parsec). Experimental results show that the prediction model achieves good accuracy, which is maximum 5.6Â % error rate.

KW - energy efficiency

KW - HPC

KW - power modeling

KW - RAPL

U2 - 10.1007/s00450-016-0325-4

DO - 10.1007/s00450-016-0325-4

M3 - Article

VL - 31

SP - 207

EP - 214

JO - SICS Software-Intensive Cyber-Physical Systems

JF - SICS Software-Intensive Cyber-Physical Systems

SN - 2524-8529

IS - 4

ER -