Adagio: Making DVS practical for complex HPC applications

Barry Rountree, David K Lowenthal, Bronis R. De Supinski, Martin Schulz, Vincent W. Freeh, Tyler Bletsch

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

179 Citations (Scopus)

Abstract

Power and energy are first-order design constraints in high performance computing. Current research using dynamic voltage scaling (DVS) relies on trading increased execution time for energy savings, which is unacceptable for most high performance computing applications. We present Adagio, a novel runtime system that makes DVS practical for complex, real-world scientific applications by incurring only negligible delay while achieving signifi-cant energy savings. Adagio improves and extends previous stateof-the-art algorithms by combining the lessons learned from static energy-reducing CPU scheduling with a novel runtime mechanism for slack prediction. We present results using Adagio for two realworld programs, UMT2K and ParaDiS, along with the NAS Parallel Benchmark suite. While requiring no modification to the application source code, Adagio provides total system energy savings of 8% and 20% for UMT2K and ParaDiS, respectively, with less than 1% increase in execution time.

Original languageEnglish (US)
Title of host publicationProceedings of the International Conference on Supercomputing
Pages460-469
Number of pages10
DOIs
StatePublished - 2009
Event23rd International Conference on Supercomputing, ICS'09 - Yorktown Heights, NY, United States
Duration: Jun 8 2009Jun 12 2009

Other

Other23rd International Conference on Supercomputing, ICS'09
CountryUnited States
CityYorktown Heights, NY
Period6/8/096/12/09

Fingerprint

Energy conservation
Program processors
Scheduling
Voltage scaling

Keywords

  • DVFS
  • DVS
  • Energy
  • MPI
  • Runtime

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Rountree, B., Lowenthal, D. K., De Supinski, B. R., Schulz, M., Freeh, V. W., & Bletsch, T. (2009). Adagio: Making DVS practical for complex HPC applications. In Proceedings of the International Conference on Supercomputing (pp. 460-469). [1542340] https://doi.org/10.1145/1542275.1542340

Adagio : Making DVS practical for complex HPC applications. / Rountree, Barry; Lowenthal, David K; De Supinski, Bronis R.; Schulz, Martin; Freeh, Vincent W.; Bletsch, Tyler.

Proceedings of the International Conference on Supercomputing. 2009. p. 460-469 1542340.

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

Rountree, B, Lowenthal, DK, De Supinski, BR, Schulz, M, Freeh, VW & Bletsch, T 2009, Adagio: Making DVS practical for complex HPC applications. in Proceedings of the International Conference on Supercomputing., 1542340, pp. 460-469, 23rd International Conference on Supercomputing, ICS'09, Yorktown Heights, NY, United States, 6/8/09. https://doi.org/10.1145/1542275.1542340
Rountree B, Lowenthal DK, De Supinski BR, Schulz M, Freeh VW, Bletsch T. Adagio: Making DVS practical for complex HPC applications. In Proceedings of the International Conference on Supercomputing. 2009. p. 460-469. 1542340 https://doi.org/10.1145/1542275.1542340
Rountree, Barry ; Lowenthal, David K ; De Supinski, Bronis R. ; Schulz, Martin ; Freeh, Vincent W. ; Bletsch, Tyler. / Adagio : Making DVS practical for complex HPC applications. Proceedings of the International Conference on Supercomputing. 2009. pp. 460-469
@inproceedings{744b1c3c874c4b8faf6c5a0e1ca983d3,
title = "Adagio: Making DVS practical for complex HPC applications",
abstract = "Power and energy are first-order design constraints in high performance computing. Current research using dynamic voltage scaling (DVS) relies on trading increased execution time for energy savings, which is unacceptable for most high performance computing applications. We present Adagio, a novel runtime system that makes DVS practical for complex, real-world scientific applications by incurring only negligible delay while achieving signifi-cant energy savings. Adagio improves and extends previous stateof-the-art algorithms by combining the lessons learned from static energy-reducing CPU scheduling with a novel runtime mechanism for slack prediction. We present results using Adagio for two realworld programs, UMT2K and ParaDiS, along with the NAS Parallel Benchmark suite. While requiring no modification to the application source code, Adagio provides total system energy savings of 8{\%} and 20{\%} for UMT2K and ParaDiS, respectively, with less than 1{\%} increase in execution time.",
keywords = "DVFS, DVS, Energy, MPI, Runtime",
author = "Barry Rountree and Lowenthal, {David K} and {De Supinski}, {Bronis R.} and Martin Schulz and Freeh, {Vincent W.} and Tyler Bletsch",
year = "2009",
doi = "10.1145/1542275.1542340",
language = "English (US)",
isbn = "9781605584980",
pages = "460--469",
booktitle = "Proceedings of the International Conference on Supercomputing",

}

TY - GEN

T1 - Adagio

T2 - Making DVS practical for complex HPC applications

AU - Rountree, Barry

AU - Lowenthal, David K

AU - De Supinski, Bronis R.

AU - Schulz, Martin

AU - Freeh, Vincent W.

AU - Bletsch, Tyler

PY - 2009

Y1 - 2009

N2 - Power and energy are first-order design constraints in high performance computing. Current research using dynamic voltage scaling (DVS) relies on trading increased execution time for energy savings, which is unacceptable for most high performance computing applications. We present Adagio, a novel runtime system that makes DVS practical for complex, real-world scientific applications by incurring only negligible delay while achieving signifi-cant energy savings. Adagio improves and extends previous stateof-the-art algorithms by combining the lessons learned from static energy-reducing CPU scheduling with a novel runtime mechanism for slack prediction. We present results using Adagio for two realworld programs, UMT2K and ParaDiS, along with the NAS Parallel Benchmark suite. While requiring no modification to the application source code, Adagio provides total system energy savings of 8% and 20% for UMT2K and ParaDiS, respectively, with less than 1% increase in execution time.

AB - Power and energy are first-order design constraints in high performance computing. Current research using dynamic voltage scaling (DVS) relies on trading increased execution time for energy savings, which is unacceptable for most high performance computing applications. We present Adagio, a novel runtime system that makes DVS practical for complex, real-world scientific applications by incurring only negligible delay while achieving signifi-cant energy savings. Adagio improves and extends previous stateof-the-art algorithms by combining the lessons learned from static energy-reducing CPU scheduling with a novel runtime mechanism for slack prediction. We present results using Adagio for two realworld programs, UMT2K and ParaDiS, along with the NAS Parallel Benchmark suite. While requiring no modification to the application source code, Adagio provides total system energy savings of 8% and 20% for UMT2K and ParaDiS, respectively, with less than 1% increase in execution time.

KW - DVFS

KW - DVS

KW - Energy

KW - MPI

KW - Runtime

UR - http://www.scopus.com/inward/record.url?scp=70449728146&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70449728146&partnerID=8YFLogxK

U2 - 10.1145/1542275.1542340

DO - 10.1145/1542275.1542340

M3 - Conference contribution

AN - SCOPUS:70449728146

SN - 9781605584980

SP - 460

EP - 469

BT - Proceedings of the International Conference on Supercomputing

ER -