Reducing power with performance constraints for parallel sparse applications

G. Chen, K. Malkowski, Mahmut Kandemir, P. Raghavan

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

40 Citations (Scopus)

Abstract

Sparse and irregular computations constitute a large fraction of applications in the data-intensive scientific domain. While every effort is made to balance the computational workload in such computations across parallel processors, achieving sustained near machine-peak performance with close-to-ideal load balanced computation-to-processor mapping is inherently difficult. As a result, most of the time, the loads assigned to parallel processors can exhibit significant variations. While there have been numerous past efforts that study this imbalance from the performance viewpoint, to our knowledge, no prior study has considered exploiting the imbalance for reducing power consumption during execution. Power consumption in large-scale clusters of workstations is becoming a critical issue as noted by several recent research papers from both industry and academia. Focusing on sparse matrix computations in which underlying parallel computations and data dependencies can be represented by trees, this paper proposes schemes that save power through voltage/frequency scaling. Our goal is to reduce overall energy consumption by scaling the voltages/frequencies of those processors that are not in the critical path; i.e., our approach is oriented towards saving power without incurring performance penalties.

Original languageEnglish (US)
Title of host publicationProceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005
DOIs
StatePublished - Dec 1 2005
Event19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005 - Denver, CO, United States
Duration: Apr 4 2005Apr 8 2005

Publication series

NameProceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005
Volume2005

Other

Other19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005
CountryUnited States
CityDenver, CO
Period4/4/054/8/05

Fingerprint

Electric power utilization
Electric potential
Energy utilization
Industry

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

Chen, G., Malkowski, K., Kandemir, M., & Raghavan, P. (2005). Reducing power with performance constraints for parallel sparse applications. In Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005 [1420150] (Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005; Vol. 2005). https://doi.org/10.1109/IPDPS.2005.378
Chen, G. ; Malkowski, K. ; Kandemir, Mahmut ; Raghavan, P. / Reducing power with performance constraints for parallel sparse applications. Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005. 2005. (Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005).
@inproceedings{6f688f6b373b4269b91142ab5e40e566,
title = "Reducing power with performance constraints for parallel sparse applications",
abstract = "Sparse and irregular computations constitute a large fraction of applications in the data-intensive scientific domain. While every effort is made to balance the computational workload in such computations across parallel processors, achieving sustained near machine-peak performance with close-to-ideal load balanced computation-to-processor mapping is inherently difficult. As a result, most of the time, the loads assigned to parallel processors can exhibit significant variations. While there have been numerous past efforts that study this imbalance from the performance viewpoint, to our knowledge, no prior study has considered exploiting the imbalance for reducing power consumption during execution. Power consumption in large-scale clusters of workstations is becoming a critical issue as noted by several recent research papers from both industry and academia. Focusing on sparse matrix computations in which underlying parallel computations and data dependencies can be represented by trees, this paper proposes schemes that save power through voltage/frequency scaling. Our goal is to reduce overall energy consumption by scaling the voltages/frequencies of those processors that are not in the critical path; i.e., our approach is oriented towards saving power without incurring performance penalties.",
author = "G. Chen and K. Malkowski and Mahmut Kandemir and P. Raghavan",
year = "2005",
month = "12",
day = "1",
doi = "10.1109/IPDPS.2005.378",
language = "English (US)",
isbn = "0769523129",
series = "Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005",
booktitle = "Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005",

}

Chen, G, Malkowski, K, Kandemir, M & Raghavan, P 2005, Reducing power with performance constraints for parallel sparse applications. in Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005., 1420150, Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005, vol. 2005, 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005, Denver, CO, United States, 4/4/05. https://doi.org/10.1109/IPDPS.2005.378

Reducing power with performance constraints for parallel sparse applications. / Chen, G.; Malkowski, K.; Kandemir, Mahmut; Raghavan, P.

Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005. 2005. 1420150 (Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005; Vol. 2005).

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

TY - GEN

T1 - Reducing power with performance constraints for parallel sparse applications

AU - Chen, G.

AU - Malkowski, K.

AU - Kandemir, Mahmut

AU - Raghavan, P.

PY - 2005/12/1

Y1 - 2005/12/1

N2 - Sparse and irregular computations constitute a large fraction of applications in the data-intensive scientific domain. While every effort is made to balance the computational workload in such computations across parallel processors, achieving sustained near machine-peak performance with close-to-ideal load balanced computation-to-processor mapping is inherently difficult. As a result, most of the time, the loads assigned to parallel processors can exhibit significant variations. While there have been numerous past efforts that study this imbalance from the performance viewpoint, to our knowledge, no prior study has considered exploiting the imbalance for reducing power consumption during execution. Power consumption in large-scale clusters of workstations is becoming a critical issue as noted by several recent research papers from both industry and academia. Focusing on sparse matrix computations in which underlying parallel computations and data dependencies can be represented by trees, this paper proposes schemes that save power through voltage/frequency scaling. Our goal is to reduce overall energy consumption by scaling the voltages/frequencies of those processors that are not in the critical path; i.e., our approach is oriented towards saving power without incurring performance penalties.

AB - Sparse and irregular computations constitute a large fraction of applications in the data-intensive scientific domain. While every effort is made to balance the computational workload in such computations across parallel processors, achieving sustained near machine-peak performance with close-to-ideal load balanced computation-to-processor mapping is inherently difficult. As a result, most of the time, the loads assigned to parallel processors can exhibit significant variations. While there have been numerous past efforts that study this imbalance from the performance viewpoint, to our knowledge, no prior study has considered exploiting the imbalance for reducing power consumption during execution. Power consumption in large-scale clusters of workstations is becoming a critical issue as noted by several recent research papers from both industry and academia. Focusing on sparse matrix computations in which underlying parallel computations and data dependencies can be represented by trees, this paper proposes schemes that save power through voltage/frequency scaling. Our goal is to reduce overall energy consumption by scaling the voltages/frequencies of those processors that are not in the critical path; i.e., our approach is oriented towards saving power without incurring performance penalties.

UR - http://www.scopus.com/inward/record.url?scp=33746318690&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33746318690&partnerID=8YFLogxK

U2 - 10.1109/IPDPS.2005.378

DO - 10.1109/IPDPS.2005.378

M3 - Conference contribution

SN - 0769523129

SN - 0769523129

SN - 9780769523125

T3 - Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005

BT - Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005

ER -

Chen G, Malkowski K, Kandemir M, Raghavan P. Reducing power with performance constraints for parallel sparse applications. In Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005. 2005. 1420150. (Proceedings - 19th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2005). https://doi.org/10.1109/IPDPS.2005.378