Analyzing the energy-time trade-off in high-performance computing applications

Vincent W. Freeh, David K Lowenthal, Feng Pan, Nandini Kappiah, Rob Springer, Barry L. Rountree, Mark E. Femal

Research output: Contribution to journalArticle

127 Citations (Scopus)

Abstract

Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. One approach to lowering energy and power is to use high-performance cluster nodes that have several power-performance states so that the energy-time trade-off can be dynamically adjusted. This paper analyzes the energy-time trade-off of a wide range of applications serial and parallel on a power-scalable cluster. We use a cluster of frequency and voltage-scalable AMD-64 nodes, each equipped with a power meter. We study the effects of memory and communication bottlenecks via direct measurement of time and energy. We also investigate metrics that can, at runtime, predict when each type of bottleneck occurs. Our results show that, for programs that have a memory or communication bottleneck, a power-scalable cluster can save significant energy with only a small time penalty. Furthermore, we find that, for some programs, it is possible to both consume less energy and execute in less time by increasing the number of nodes while reducing the frequency-voltage setting of each node.

Original languageEnglish (US)
Pages (from-to)835-848
Number of pages14
JournalIEEE Transactions on Parallel and Distributed Systems
Volume18
Issue number6
DOIs
StatePublished - Jun 2007
Externally publishedYes

Fingerprint

High Performance
Trade-offs
Data storage equipment
Computing
Communication
Electric potential
Energy
Electric power utilization
Energy utilization
Vertex of a graph
Voltage
Power Consumption
Energy Consumption
Penalty
Metric
Predict
Range of data

Keywords

  • High-performance computing
  • Power-aware computing

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Theoretical Computer Science
  • Computational Theory and Mathematics

Cite this

Analyzing the energy-time trade-off in high-performance computing applications. / Freeh, Vincent W.; Lowenthal, David K; Pan, Feng; Kappiah, Nandini; Springer, Rob; Rountree, Barry L.; Femal, Mark E.

In: IEEE Transactions on Parallel and Distributed Systems, Vol. 18, No. 6, 06.2007, p. 835-848.

Research output: Contribution to journalArticle

Freeh, Vincent W. ; Lowenthal, David K ; Pan, Feng ; Kappiah, Nandini ; Springer, Rob ; Rountree, Barry L. ; Femal, Mark E. / Analyzing the energy-time trade-off in high-performance computing applications. In: IEEE Transactions on Parallel and Distributed Systems. 2007 ; Vol. 18, No. 6. pp. 835-848.
@article{5ab9ffb563e8428180bf23b4c606fd57,
title = "Analyzing the energy-time trade-off in high-performance computing applications",
abstract = "Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. One approach to lowering energy and power is to use high-performance cluster nodes that have several power-performance states so that the energy-time trade-off can be dynamically adjusted. This paper analyzes the energy-time trade-off of a wide range of applications serial and parallel on a power-scalable cluster. We use a cluster of frequency and voltage-scalable AMD-64 nodes, each equipped with a power meter. We study the effects of memory and communication bottlenecks via direct measurement of time and energy. We also investigate metrics that can, at runtime, predict when each type of bottleneck occurs. Our results show that, for programs that have a memory or communication bottleneck, a power-scalable cluster can save significant energy with only a small time penalty. Furthermore, we find that, for some programs, it is possible to both consume less energy and execute in less time by increasing the number of nodes while reducing the frequency-voltage setting of each node.",
keywords = "High-performance computing, Power-aware computing",
author = "Freeh, {Vincent W.} and Lowenthal, {David K} and Feng Pan and Nandini Kappiah and Rob Springer and Rountree, {Barry L.} and Femal, {Mark E.}",
year = "2007",
month = "6",
doi = "10.1109/TPDS.2007.1026",
language = "English (US)",
volume = "18",
pages = "835--848",
journal = "IEEE Transactions on Parallel and Distributed Systems",
issn = "1045-9219",
publisher = "IEEE Computer Society",
number = "6",

}

TY - JOUR

T1 - Analyzing the energy-time trade-off in high-performance computing applications

AU - Freeh, Vincent W.

AU - Lowenthal, David K

AU - Pan, Feng

AU - Kappiah, Nandini

AU - Springer, Rob

AU - Rountree, Barry L.

AU - Femal, Mark E.

PY - 2007/6

Y1 - 2007/6

N2 - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. One approach to lowering energy and power is to use high-performance cluster nodes that have several power-performance states so that the energy-time trade-off can be dynamically adjusted. This paper analyzes the energy-time trade-off of a wide range of applications serial and parallel on a power-scalable cluster. We use a cluster of frequency and voltage-scalable AMD-64 nodes, each equipped with a power meter. We study the effects of memory and communication bottlenecks via direct measurement of time and energy. We also investigate metrics that can, at runtime, predict when each type of bottleneck occurs. Our results show that, for programs that have a memory or communication bottleneck, a power-scalable cluster can save significant energy with only a small time penalty. Furthermore, we find that, for some programs, it is possible to both consume less energy and execute in less time by increasing the number of nodes while reducing the frequency-voltage setting of each node.

AB - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. One approach to lowering energy and power is to use high-performance cluster nodes that have several power-performance states so that the energy-time trade-off can be dynamically adjusted. This paper analyzes the energy-time trade-off of a wide range of applications serial and parallel on a power-scalable cluster. We use a cluster of frequency and voltage-scalable AMD-64 nodes, each equipped with a power meter. We study the effects of memory and communication bottlenecks via direct measurement of time and energy. We also investigate metrics that can, at runtime, predict when each type of bottleneck occurs. Our results show that, for programs that have a memory or communication bottleneck, a power-scalable cluster can save significant energy with only a small time penalty. Furthermore, we find that, for some programs, it is possible to both consume less energy and execute in less time by increasing the number of nodes while reducing the frequency-voltage setting of each node.

KW - High-performance computing

KW - Power-aware computing

UR - http://www.scopus.com/inward/record.url?scp=34248638757&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34248638757&partnerID=8YFLogxK

U2 - 10.1109/TPDS.2007.1026

DO - 10.1109/TPDS.2007.1026

M3 - Article

VL - 18

SP - 835

EP - 848

JO - IEEE Transactions on Parallel and Distributed Systems

JF - IEEE Transactions on Parallel and Distributed Systems

SN - 1045-9219

IS - 6

ER -