Adaptive, transparent CPU scaling algorithms leveraging inter-node MPI communication regions

Min Yeol Lim, Vincent W. Freeh, David K Lowenthal

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Because the CPU is often the major power consumer, some microprocessors allow frequency and voltage scaling, which enables a system to efficiently reduce CPU performance and power. When the CPU is not on the critical path, such dynamic frequency and voltage scaling can produce significant energy savings with little performance penalty. This paper presents an MPI runtime system that dynamically reduces CPU frequency and voltage during communication phases in MPI programs. It dynamically identifies such phases and, without a priori knowledge, selects the CPU frequency in order to minimize energy-delay product. All analysis and subsequent frequency and voltage scaling is within MPI and so is entirely transparent to the application. This means that the large number of existing MPI programs, as well as new ones being developed, can use our system without modification. Results show that the median reduction in energy-delay product for twelve benchmarks is 8%, the median energy reduction is 11%, and the median increase in execution time increase is only 2%.

Original languageEnglish (US)
Pages (from-to)667-683
Number of pages17
JournalParallel Computing
Volume37
Issue number10-11
DOIs
StatePublished - Oct 2011

Fingerprint

Program processors
Scaling
Voltage
Communication
Vertex of a graph
Energy
Runtime Systems
Critical Path
Microprocessor
Energy Saving
Execution Time
Power Consumption
Energy Consumption
Penalty
Microprocessor chips
Energy conservation
Electric power utilization
Energy utilization
High Performance
Benchmark

Keywords

  • Message passing interface (MPI)
  • Power-aware computing

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Software
  • Hardware and Architecture
  • Artificial Intelligence
  • Computer Graphics and Computer-Aided Design
  • Theoretical Computer Science

Cite this

Adaptive, transparent CPU scaling algorithms leveraging inter-node MPI communication regions. / Lim, Min Yeol; Freeh, Vincent W.; Lowenthal, David K.

In: Parallel Computing, Vol. 37, No. 10-11, 10.2011, p. 667-683.

Research output: Contribution to journalArticle

@article{3c9e2adbf2ba43be8e457b5db548b104,
title = "Adaptive, transparent CPU scaling algorithms leveraging inter-node MPI communication regions",
abstract = "Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Because the CPU is often the major power consumer, some microprocessors allow frequency and voltage scaling, which enables a system to efficiently reduce CPU performance and power. When the CPU is not on the critical path, such dynamic frequency and voltage scaling can produce significant energy savings with little performance penalty. This paper presents an MPI runtime system that dynamically reduces CPU frequency and voltage during communication phases in MPI programs. It dynamically identifies such phases and, without a priori knowledge, selects the CPU frequency in order to minimize energy-delay product. All analysis and subsequent frequency and voltage scaling is within MPI and so is entirely transparent to the application. This means that the large number of existing MPI programs, as well as new ones being developed, can use our system without modification. Results show that the median reduction in energy-delay product for twelve benchmarks is 8{\%}, the median energy reduction is 11{\%}, and the median increase in execution time increase is only 2{\%}.",
keywords = "Message passing interface (MPI), Power-aware computing",
author = "Lim, {Min Yeol} and Freeh, {Vincent W.} and Lowenthal, {David K}",
year = "2011",
month = "10",
doi = "10.1016/j.parco.2011.07.001",
language = "English (US)",
volume = "37",
pages = "667--683",
journal = "Parallel Computing",
issn = "0167-8191",
publisher = "Elsevier",
number = "10-11",

}

TY - JOUR

T1 - Adaptive, transparent CPU scaling algorithms leveraging inter-node MPI communication regions

AU - Lim, Min Yeol

AU - Freeh, Vincent W.

AU - Lowenthal, David K

PY - 2011/10

Y1 - 2011/10

N2 - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Because the CPU is often the major power consumer, some microprocessors allow frequency and voltage scaling, which enables a system to efficiently reduce CPU performance and power. When the CPU is not on the critical path, such dynamic frequency and voltage scaling can produce significant energy savings with little performance penalty. This paper presents an MPI runtime system that dynamically reduces CPU frequency and voltage during communication phases in MPI programs. It dynamically identifies such phases and, without a priori knowledge, selects the CPU frequency in order to minimize energy-delay product. All analysis and subsequent frequency and voltage scaling is within MPI and so is entirely transparent to the application. This means that the large number of existing MPI programs, as well as new ones being developed, can use our system without modification. Results show that the median reduction in energy-delay product for twelve benchmarks is 8%, the median energy reduction is 11%, and the median increase in execution time increase is only 2%.

AB - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Because the CPU is often the major power consumer, some microprocessors allow frequency and voltage scaling, which enables a system to efficiently reduce CPU performance and power. When the CPU is not on the critical path, such dynamic frequency and voltage scaling can produce significant energy savings with little performance penalty. This paper presents an MPI runtime system that dynamically reduces CPU frequency and voltage during communication phases in MPI programs. It dynamically identifies such phases and, without a priori knowledge, selects the CPU frequency in order to minimize energy-delay product. All analysis and subsequent frequency and voltage scaling is within MPI and so is entirely transparent to the application. This means that the large number of existing MPI programs, as well as new ones being developed, can use our system without modification. Results show that the median reduction in energy-delay product for twelve benchmarks is 8%, the median energy reduction is 11%, and the median increase in execution time increase is only 2%.

KW - Message passing interface (MPI)

KW - Power-aware computing

UR - http://www.scopus.com/inward/record.url?scp=80051860489&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80051860489&partnerID=8YFLogxK

U2 - 10.1016/j.parco.2011.07.001

DO - 10.1016/j.parco.2011.07.001

M3 - Article

AN - SCOPUS:80051860489

VL - 37

SP - 667

EP - 683

JO - Parallel Computing

JF - Parallel Computing

SN - 0167-8191

IS - 10-11

ER -