A power-measurement methodology for large-scale, high-performance computing

Thomas R.W. Scogland, Craig Philip Steffen, Torsten Wilde, Florent Parent, Susan Coghlan, Natalie Bates, Wu Chun Feng, Erich Strohmaier

Research output: Contribution to conferencePaper

Abstract

Improvement in the energy efficiency of supercomputers can be accelerated by improving the quality and comparability of efficiency measurements. The ability to generate accurate measurements at extreme scale are just now emerging. The realization of system-level measurement capabilities can be accelerated with a commonly adopted and high quality measurement methodology for use while running a workload, typically a benchmark. This paper describes a methodology that has been developed collaboratively through the Energy Efficient HPC Working Group to support architectural analysis and comparative measurements for rankings, such as the Top500 and Green500. To support measurements with varying amounts of effort and equipment required we present three distinct levels of measurement, which provide increasing levels of accuracy. Level 1 is similar to the Green500 run rules today, a single average power measurement extrapolated from a subset of a machine. Level 2 is more comprehensive, but still widely achievable. Level 3 is the most rigorous of the three methodologies but is only possible at a few sites. However, the Level 3 methodology generates a high quality result that exposes details that the other methodologies may miss. In addition, we present case studies from the Leibniz Supercomputing Centre (LRZ), Argonne National Laboratory (ANL) and Calcul Québec Université Laval that explore the benefits and difficulties of gathering high quality, system-level measurements on large-scale machines.

Original languageEnglish (US)
Pages149-159
Number of pages11
DOIs
StatePublished - Jan 1 2014
Event5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014 - Dublin, Ireland
Duration: Mar 22 2014Mar 26 2014

Other

Other5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014
CountryIreland
CityDublin
Period3/22/143/26/14

Fingerprint

Level measurement
Supercomputers
Energy efficiency

Keywords

  • Datacenter
  • Green500
  • Highperformance computing
  • Power-measurement methodology
  • Top500

ASJC Scopus subject areas

  • Software

Cite this

Scogland, T. R. W., Steffen, C. P., Wilde, T., Parent, F., Coghlan, S., Bates, N., ... Strohmaier, E. (2014). A power-measurement methodology for large-scale, high-performance computing. 149-159. Paper presented at 5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014, Dublin, Ireland. https://doi.org/10.1145/2568088.2576795

A power-measurement methodology for large-scale, high-performance computing. / Scogland, Thomas R.W.; Steffen, Craig Philip; Wilde, Torsten; Parent, Florent; Coghlan, Susan; Bates, Natalie; Feng, Wu Chun; Strohmaier, Erich.

2014. 149-159 Paper presented at 5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014, Dublin, Ireland.

Research output: Contribution to conferencePaper

Scogland, TRW, Steffen, CP, Wilde, T, Parent, F, Coghlan, S, Bates, N, Feng, WC & Strohmaier, E 2014, 'A power-measurement methodology for large-scale, high-performance computing' Paper presented at 5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014, Dublin, Ireland, 3/22/14 - 3/26/14, pp. 149-159. https://doi.org/10.1145/2568088.2576795
Scogland TRW, Steffen CP, Wilde T, Parent F, Coghlan S, Bates N et al. A power-measurement methodology for large-scale, high-performance computing. 2014. Paper presented at 5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014, Dublin, Ireland. https://doi.org/10.1145/2568088.2576795
Scogland, Thomas R.W. ; Steffen, Craig Philip ; Wilde, Torsten ; Parent, Florent ; Coghlan, Susan ; Bates, Natalie ; Feng, Wu Chun ; Strohmaier, Erich. / A power-measurement methodology for large-scale, high-performance computing. Paper presented at 5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014, Dublin, Ireland.11 p.
@conference{f46babb46c4d43b4b915f7f5ad38d512,
title = "A power-measurement methodology for large-scale, high-performance computing",
abstract = "Improvement in the energy efficiency of supercomputers can be accelerated by improving the quality and comparability of efficiency measurements. The ability to generate accurate measurements at extreme scale are just now emerging. The realization of system-level measurement capabilities can be accelerated with a commonly adopted and high quality measurement methodology for use while running a workload, typically a benchmark. This paper describes a methodology that has been developed collaboratively through the Energy Efficient HPC Working Group to support architectural analysis and comparative measurements for rankings, such as the Top500 and Green500. To support measurements with varying amounts of effort and equipment required we present three distinct levels of measurement, which provide increasing levels of accuracy. Level 1 is similar to the Green500 run rules today, a single average power measurement extrapolated from a subset of a machine. Level 2 is more comprehensive, but still widely achievable. Level 3 is the most rigorous of the three methodologies but is only possible at a few sites. However, the Level 3 methodology generates a high quality result that exposes details that the other methodologies may miss. In addition, we present case studies from the Leibniz Supercomputing Centre (LRZ), Argonne National Laboratory (ANL) and Calcul Qu{\'e}bec Universit{\'e} Laval that explore the benefits and difficulties of gathering high quality, system-level measurements on large-scale machines.",
keywords = "Datacenter, Green500, Highperformance computing, Power-measurement methodology, Top500",
author = "Scogland, {Thomas R.W.} and Steffen, {Craig Philip} and Torsten Wilde and Florent Parent and Susan Coghlan and Natalie Bates and Feng, {Wu Chun} and Erich Strohmaier",
year = "2014",
month = "1",
day = "1",
doi = "10.1145/2568088.2576795",
language = "English (US)",
pages = "149--159",
note = "5th ACM/SPEC International Conference on Performance Engineering, ICPE 2014 ; Conference date: 22-03-2014 Through 26-03-2014",

}

TY - CONF

T1 - A power-measurement methodology for large-scale, high-performance computing

AU - Scogland, Thomas R.W.

AU - Steffen, Craig Philip

AU - Wilde, Torsten

AU - Parent, Florent

AU - Coghlan, Susan

AU - Bates, Natalie

AU - Feng, Wu Chun

AU - Strohmaier, Erich

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Improvement in the energy efficiency of supercomputers can be accelerated by improving the quality and comparability of efficiency measurements. The ability to generate accurate measurements at extreme scale are just now emerging. The realization of system-level measurement capabilities can be accelerated with a commonly adopted and high quality measurement methodology for use while running a workload, typically a benchmark. This paper describes a methodology that has been developed collaboratively through the Energy Efficient HPC Working Group to support architectural analysis and comparative measurements for rankings, such as the Top500 and Green500. To support measurements with varying amounts of effort and equipment required we present three distinct levels of measurement, which provide increasing levels of accuracy. Level 1 is similar to the Green500 run rules today, a single average power measurement extrapolated from a subset of a machine. Level 2 is more comprehensive, but still widely achievable. Level 3 is the most rigorous of the three methodologies but is only possible at a few sites. However, the Level 3 methodology generates a high quality result that exposes details that the other methodologies may miss. In addition, we present case studies from the Leibniz Supercomputing Centre (LRZ), Argonne National Laboratory (ANL) and Calcul Québec Université Laval that explore the benefits and difficulties of gathering high quality, system-level measurements on large-scale machines.

AB - Improvement in the energy efficiency of supercomputers can be accelerated by improving the quality and comparability of efficiency measurements. The ability to generate accurate measurements at extreme scale are just now emerging. The realization of system-level measurement capabilities can be accelerated with a commonly adopted and high quality measurement methodology for use while running a workload, typically a benchmark. This paper describes a methodology that has been developed collaboratively through the Energy Efficient HPC Working Group to support architectural analysis and comparative measurements for rankings, such as the Top500 and Green500. To support measurements with varying amounts of effort and equipment required we present three distinct levels of measurement, which provide increasing levels of accuracy. Level 1 is similar to the Green500 run rules today, a single average power measurement extrapolated from a subset of a machine. Level 2 is more comprehensive, but still widely achievable. Level 3 is the most rigorous of the three methodologies but is only possible at a few sites. However, the Level 3 methodology generates a high quality result that exposes details that the other methodologies may miss. In addition, we present case studies from the Leibniz Supercomputing Centre (LRZ), Argonne National Laboratory (ANL) and Calcul Québec Université Laval that explore the benefits and difficulties of gathering high quality, system-level measurements on large-scale machines.

KW - Datacenter

KW - Green500

KW - Highperformance computing

KW - Power-measurement methodology

KW - Top500

UR - http://www.scopus.com/inward/record.url?scp=84899696900&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84899696900&partnerID=8YFLogxK

U2 - 10.1145/2568088.2576795

DO - 10.1145/2568088.2576795

M3 - Paper

AN - SCOPUS:84899696900

SP - 149

EP - 159

ER -