Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments

Yectli A. Huerta; Brent Swartz; David J Lilja

doi:10.1109/IISWC.2017.8167766

Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments

Yectli A. Huerta, Brent Swartz, David J Lilja

Electrical and Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

In a closely coupled heterogeneous computing system the work is shared amongst all available computing resources. One challenge is to find an optimal division of work between the two or more very different kinds of processing units, each with their own optimal settings. We show that through the use of statistical techniques, a systematic search of the parameter space can be conducted. These techniques can be applied to variables that are categorical or continuous in nature and do not rely on the standard assumptions of linear models, mainly that the response variable can be described as a linear combination of the regression coefficients. Our search technique, when applied to the HPL benchmark, resulted in a performance gain of 14.5% over previously reported results.

Original language	English (US)
Title of host publication	Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	118-119
Number of pages	2
ISBN (Electronic)	9781538612323
DOIs	https://doi.org/10.1109/IISWC.2017.8167766
State	Published - Dec 5 2017
Event	2017 IEEE International Symposium on Workload Characterization, IISWC 2017 - Seattle, United States Duration: Oct 1 2017 → Oct 3 2017

Publication series

Name	Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017
Volume	2017-January

Other

Other	2017 IEEE International Symposium on Workload Characterization, IISWC 2017
Country/Territory	United States
City	Seattle
Period	10/1/17 → 10/3/17

Bibliographical note

Funding Information:
ACKNOWLEDGMENT We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF.

Funding Information:
We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF.

Publisher Copyright:
© 2017 IEEE.

Access

10.1109/IISWC.2017.8167766

OpenUrl availability

Full text

Cite this

Huerta, Y. A., Swartz, B., & Lilja, D. J. (2017). Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments. In Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017 (pp. 118-119). (Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017; Vol. 2017-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IISWC.2017.8167766

Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments. / Huerta, Yectli A.; Swartz, Brent; Lilja, David J.
Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 118-119 (Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017; Vol. 2017-January).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Huerta, YA, Swartz, B & Lilja, DJ 2017, Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments. in Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017. Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017, vol. 2017-January, Institute of Electrical and Electronics Engineers Inc., pp. 118-119, 2017 IEEE International Symposium on Workload Characterization, IISWC 2017, Seattle, United States, 10/1/17. https://doi.org/10.1109/IISWC.2017.8167766

Huerta YA, Swartz B, Lilja DJ. Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments. In Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 118-119. (Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017). doi: 10.1109/IISWC.2017.8167766

Huerta, Yectli A. ; Swartz, Brent ; Lilja, David J. / Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments. Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 118-119 (Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017).

@inproceedings{f4bf64c1906e4214851a513f09cfaf73,

title = "Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments",

abstract = "In a closely coupled heterogeneous computing system the work is shared amongst all available computing resources. One challenge is to find an optimal division of work between the two or more very different kinds of processing units, each with their own optimal settings. We show that through the use of statistical techniques, a systematic search of the parameter space can be conducted. These techniques can be applied to variables that are categorical or continuous in nature and do not rely on the standard assumptions of linear models, mainly that the response variable can be described as a linear combination of the regression coefficients. Our search technique, when applied to the HPL benchmark, resulted in a performance gain of 14.5% over previously reported results.",

author = "Huerta, {Yectli A.} and Brent Swartz and Lilja, {David J}",

note = "Funding Information: ACKNOWLEDGMENT We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. Funding Information: We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. Publisher Copyright: {\textcopyright} 2017 IEEE.; 2017 IEEE International Symposium on Workload Characterization, IISWC 2017 ; Conference date: 01-10-2017 Through 03-10-2017",

year = "2017",

month = dec,

day = "5",

doi = "10.1109/IISWC.2017.8167766",

language = "English (US)",

series = "Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "118--119",

booktitle = "Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017",

}

TY - GEN

T1 - Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments

AU - Huerta, Yectli A.

AU - Swartz, Brent

AU - Lilja, David J

N1 - Funding Information: ACKNOWLEDGMENT We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. Funding Information: We thank the Minnesota Supercomputing Institute for the use of their computational resources. This work was supported in part by National Science Foundation grant no. CCF-1438286. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. Publisher Copyright: © 2017 IEEE.

PY - 2017/12/5

Y1 - 2017/12/5

N2 - In a closely coupled heterogeneous computing system the work is shared amongst all available computing resources. One challenge is to find an optimal division of work between the two or more very different kinds of processing units, each with their own optimal settings. We show that through the use of statistical techniques, a systematic search of the parameter space can be conducted. These techniques can be applied to variables that are categorical or continuous in nature and do not rely on the standard assumptions of linear models, mainly that the response variable can be described as a linear combination of the regression coefficients. Our search technique, when applied to the HPL benchmark, resulted in a performance gain of 14.5% over previously reported results.

AB - In a closely coupled heterogeneous computing system the work is shared amongst all available computing resources. One challenge is to find an optimal division of work between the two or more very different kinds of processing units, each with their own optimal settings. We show that through the use of statistical techniques, a systematic search of the parameter space can be conducted. These techniques can be applied to variables that are categorical or continuous in nature and do not rely on the standard assumptions of linear models, mainly that the response variable can be described as a linear combination of the regression coefficients. Our search technique, when applied to the HPL benchmark, resulted in a performance gain of 14.5% over previously reported results.

UR - http://www.scopus.com/inward/record.url?scp=85046549123&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046549123&partnerID=8YFLogxK

U2 - 10.1109/IISWC.2017.8167766

DO - 10.1109/IISWC.2017.8167766

M3 - Conference contribution

AN - SCOPUS:85046549123

T3 - Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017

SP - 118

EP - 119

BT - Proceedings of the 2017 IEEE International Symposium on Workload Characterization, IISWC 2017

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2017 IEEE International Symposium on Workload Characterization, IISWC 2017

Y2 - 1 October 2017 through 3 October 2017

ER -

Determining work partitioning on closely coupled heterogeneous computing systems using statistical design of experiments

Abstract

Publication series

Other

Bibliographical note

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this