Finding the limits of power-constrained application performance

Peter E. Bailey, Aniruddha Marathe, David K Lowenthal, Barry Rountree, Martin Schulz

Research output: Chapter in Book/Report/Conference proceedingConference contribution

24 Citations (Scopus)

Abstract

As we approach exascale systems, power is turning from an optimization goal to a critical operating constraint. With power bounds imposed by both stakeholders and the limitations of existing infrastructure, we need to develop new techniques that work with limited power to extract maximum performance. In this paper, we explore this area and provide an approach to find the theoretical upper bound of computational performance on a per-application basis in hybrid MPI + OpenMP applications. We use a linear programming (LP) formulation to optimize application schedules under various power constraints, where a schedule consists of a DVFS state and number of OpenMP threads for each section of computation between consecutive MPI calls. We also provide a more flexible mixed integer-linear (ILP) formulation and show that the resulting schedules closely match schedules from the LP formulation. Across four applications, we use our LP-derived upper bounds to show that current approaches trail optimal, power-constrained performance by up to 41.1%. This demonstrates the untapped potential of current systems, and our LP formulation provides future optimization approaches with a quantitative optimization target.

Original languageEnglish (US)
Title of host publicationInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC
PublisherIEEE Computer Society
Volume15-20-November-2015
ISBN (Print)9781450337236
DOIs
StatePublished - Nov 15 2015
EventInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC 2015 - Austin, United States
Duration: Nov 15 2015Nov 20 2015

Other

OtherInternational Conference for High Performance Computing, Networking, Storage and Analysis, SC 2015
CountryUnited States
CityAustin
Period11/15/1511/20/15

Fingerprint

Linear programming
Inductive logic programming (ILP)

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications
  • Hardware and Architecture
  • Software

Cite this

Bailey, P. E., Marathe, A., Lowenthal, D. K., Rountree, B., & Schulz, M. (2015). Finding the limits of power-constrained application performance. In International Conference for High Performance Computing, Networking, Storage and Analysis, SC (Vol. 15-20-November-2015). [a79] IEEE Computer Society. https://doi.org/10.1145/2807591.2807637

Finding the limits of power-constrained application performance. / Bailey, Peter E.; Marathe, Aniruddha; Lowenthal, David K; Rountree, Barry; Schulz, Martin.

International Conference for High Performance Computing, Networking, Storage and Analysis, SC. Vol. 15-20-November-2015 IEEE Computer Society, 2015. a79.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bailey, PE, Marathe, A, Lowenthal, DK, Rountree, B & Schulz, M 2015, Finding the limits of power-constrained application performance. in International Conference for High Performance Computing, Networking, Storage and Analysis, SC. vol. 15-20-November-2015, a79, IEEE Computer Society, International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2015, Austin, United States, 11/15/15. https://doi.org/10.1145/2807591.2807637
Bailey PE, Marathe A, Lowenthal DK, Rountree B, Schulz M. Finding the limits of power-constrained application performance. In International Conference for High Performance Computing, Networking, Storage and Analysis, SC. Vol. 15-20-November-2015. IEEE Computer Society. 2015. a79 https://doi.org/10.1145/2807591.2807637
Bailey, Peter E. ; Marathe, Aniruddha ; Lowenthal, David K ; Rountree, Barry ; Schulz, Martin. / Finding the limits of power-constrained application performance. International Conference for High Performance Computing, Networking, Storage and Analysis, SC. Vol. 15-20-November-2015 IEEE Computer Society, 2015.
@inproceedings{7d374670bc3248c1a4983c32463aaf31,
title = "Finding the limits of power-constrained application performance",
abstract = "As we approach exascale systems, power is turning from an optimization goal to a critical operating constraint. With power bounds imposed by both stakeholders and the limitations of existing infrastructure, we need to develop new techniques that work with limited power to extract maximum performance. In this paper, we explore this area and provide an approach to find the theoretical upper bound of computational performance on a per-application basis in hybrid MPI + OpenMP applications. We use a linear programming (LP) formulation to optimize application schedules under various power constraints, where a schedule consists of a DVFS state and number of OpenMP threads for each section of computation between consecutive MPI calls. We also provide a more flexible mixed integer-linear (ILP) formulation and show that the resulting schedules closely match schedules from the LP formulation. Across four applications, we use our LP-derived upper bounds to show that current approaches trail optimal, power-constrained performance by up to 41.1{\%}. This demonstrates the untapped potential of current systems, and our LP formulation provides future optimization approaches with a quantitative optimization target.",
author = "Bailey, {Peter E.} and Aniruddha Marathe and Lowenthal, {David K} and Barry Rountree and Martin Schulz",
year = "2015",
month = "11",
day = "15",
doi = "10.1145/2807591.2807637",
language = "English (US)",
isbn = "9781450337236",
volume = "15-20-November-2015",
booktitle = "International Conference for High Performance Computing, Networking, Storage and Analysis, SC",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Finding the limits of power-constrained application performance

AU - Bailey, Peter E.

AU - Marathe, Aniruddha

AU - Lowenthal, David K

AU - Rountree, Barry

AU - Schulz, Martin

PY - 2015/11/15

Y1 - 2015/11/15

N2 - As we approach exascale systems, power is turning from an optimization goal to a critical operating constraint. With power bounds imposed by both stakeholders and the limitations of existing infrastructure, we need to develop new techniques that work with limited power to extract maximum performance. In this paper, we explore this area and provide an approach to find the theoretical upper bound of computational performance on a per-application basis in hybrid MPI + OpenMP applications. We use a linear programming (LP) formulation to optimize application schedules under various power constraints, where a schedule consists of a DVFS state and number of OpenMP threads for each section of computation between consecutive MPI calls. We also provide a more flexible mixed integer-linear (ILP) formulation and show that the resulting schedules closely match schedules from the LP formulation. Across four applications, we use our LP-derived upper bounds to show that current approaches trail optimal, power-constrained performance by up to 41.1%. This demonstrates the untapped potential of current systems, and our LP formulation provides future optimization approaches with a quantitative optimization target.

AB - As we approach exascale systems, power is turning from an optimization goal to a critical operating constraint. With power bounds imposed by both stakeholders and the limitations of existing infrastructure, we need to develop new techniques that work with limited power to extract maximum performance. In this paper, we explore this area and provide an approach to find the theoretical upper bound of computational performance on a per-application basis in hybrid MPI + OpenMP applications. We use a linear programming (LP) formulation to optimize application schedules under various power constraints, where a schedule consists of a DVFS state and number of OpenMP threads for each section of computation between consecutive MPI calls. We also provide a more flexible mixed integer-linear (ILP) formulation and show that the resulting schedules closely match schedules from the LP formulation. Across four applications, we use our LP-derived upper bounds to show that current approaches trail optimal, power-constrained performance by up to 41.1%. This demonstrates the untapped potential of current systems, and our LP formulation provides future optimization approaches with a quantitative optimization target.

UR - http://www.scopus.com/inward/record.url?scp=84966549024&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84966549024&partnerID=8YFLogxK

U2 - 10.1145/2807591.2807637

DO - 10.1145/2807591.2807637

M3 - Conference contribution

AN - SCOPUS:84966549024

SN - 9781450337236

VL - 15-20-November-2015

BT - International Conference for High Performance Computing, Networking, Storage and Analysis, SC

PB - IEEE Computer Society

ER -