Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs

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

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

116 Citations (Scopus)

Abstract

Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Some microprocessors allow frequency and voltage scaling, which enables a system to reduce CPU performance and power when the CPU is not on the critical path. When properly directed, 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 performance during communication phases in MPI programs. It dynamically identifies such phases and, without profiling or training, 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 average reduction in energy-delay product over the NAS benchmark suite is 10% - -the average energy reduction is 12% while the average execution time increase is only 2.1%.

Original languageEnglish (US)
Title of host publicationProceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06
DOIs
StatePublished - 2006
Externally publishedYes

Fingerprint

Program processors
Communication
Microprocessor chips
Energy conservation
Electric power utilization
Energy utilization
Voltage scaling

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Lim, M. Y., Freeh, V. W., & Lowenthal, D. K. (2006). Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs. In Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06 [1188567] https://doi.org/10.1145/1188455.1188567

Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs. / Lim, Min Yeol; Freeh, Vincent W.; Lowenthal, David K.

Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06. 2006. 1188567.

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

Lim, MY, Freeh, VW & Lowenthal, DK 2006, Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs. in Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06., 1188567. https://doi.org/10.1145/1188455.1188567
Lim MY, Freeh VW, Lowenthal DK. Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs. In Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06. 2006. 1188567 https://doi.org/10.1145/1188455.1188567
Lim, Min Yeol ; Freeh, Vincent W. ; Lowenthal, David K. / Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs. Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06. 2006.
@inproceedings{f258f2c603764cbda144218553d5bb12,
title = "Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs",
abstract = "Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Some microprocessors allow frequency and voltage scaling, which enables a system to reduce CPU performance and power when the CPU is not on the critical path. When properly directed, 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 performance during communication phases in MPI programs. It dynamically identifies such phases and, without profiling or training, 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 average reduction in energy-delay product over the NAS benchmark suite is 10{\%} - -the average energy reduction is 12{\%} while the average execution time increase is only 2.1{\%}.",
author = "Lim, {Min Yeol} and Freeh, {Vincent W.} and Lowenthal, {David K}",
year = "2006",
doi = "10.1145/1188455.1188567",
language = "English (US)",
isbn = "0769527000",
booktitle = "Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06",

}

TY - GEN

T1 - Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs

AU - Lim, Min Yeol

AU - Freeh, Vincent W.

AU - Lowenthal, David K

PY - 2006

Y1 - 2006

N2 - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Some microprocessors allow frequency and voltage scaling, which enables a system to reduce CPU performance and power when the CPU is not on the critical path. When properly directed, 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 performance during communication phases in MPI programs. It dynamically identifies such phases and, without profiling or training, 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 average reduction in energy-delay product over the NAS benchmark suite is 10% - -the average energy reduction is 12% while the average execution time increase is only 2.1%.

AB - Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Some microprocessors allow frequency and voltage scaling, which enables a system to reduce CPU performance and power when the CPU is not on the critical path. When properly directed, 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 performance during communication phases in MPI programs. It dynamically identifies such phases and, without profiling or training, 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 average reduction in energy-delay product over the NAS benchmark suite is 10% - -the average energy reduction is 12% while the average execution time increase is only 2.1%.

UR - http://www.scopus.com/inward/record.url?scp=34548243284&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34548243284&partnerID=8YFLogxK

U2 - 10.1145/1188455.1188567

DO - 10.1145/1188455.1188567

M3 - Conference contribution

SN - 0769527000

SN - 9780769527000

BT - Proceedings of the 2006 ACM/IEEE Conference on Supercomputing, SC'06

ER -