Machine-learning-based load balancing for community ice code component in CESM

Prasanna Balaprakash; Yuri Alexeev; Sheri A. Mickelson; Sven Leyffer; Robert Jacob; Anthony Craig

doi:10.1007/978-3-319-17353-5_7

Machine-learning-based load balancing for community ice code component in CESM

Prasanna Balaprakash, Yuri Alexeev, Sheri A. Mickelson, Sven Leyffer, Robert Jacob, Anthony Craig

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

11 Scopus citations

Abstract

Load balancing scientific codes on massively parallel architectures is becoming an increasingly challenging task. In this paper, we focus on the Community Earth System Model, a widely used climate modeling code. It comprises six components each of which exhibits different scalability patterns. Previously, an analytical performance model has been used to find optimal load-balancing parameter configurations for each component. Nevertheless, for the Community Ice Code component, the analytical performance model is too restrictive to capture its scalability patterns. We therefore developed machine-learning-based load-balancing algorithm. It involves fitting a surrogate model to a small number of load-balancing configurations and their corresponding runtimes. This model is then used to find high-quality parameter configurations. Compared with the current practice of expert-knowledge-based enumeration over feasible configurations, the machine-learning-based load-balancing algorithm requires six times fewer evaluations to find the optimal configuration.

Original language	English
Title of host publication	High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers
Editors	Osni Marques, Michel Dayde, Kengo Nakajima
Publisher	Springer Verlag
Pages	79-91
Number of pages	13
ISBN (Print)	9783319173528
DOIs	https://doi.org/10.1007/978-3-319-17353-5_7
State	Published - 2015
Externally published	Yes
Event	11th International Conference on High Performance Computing for Computational Science, VECPAR 2014 - Eugene, United States Duration: Jun 30 2014 → Jul 3 2014

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	8969
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th International Conference on High Performance Computing for Computational Science, VECPAR 2014
Country/Territory	United States
City	Eugene
Period	06/30/14 → 07/3/14

Funding

This work was supported by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research, under Contract DE-AC02-06CH11357. An award of computer time was provided by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. This research used resources of the Argonne Leadership Computing Facility at Argonne National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under contract DE-AC02-06CH11357. The submitted manuscript has been created by the UChicago Argonne, LLC, Operator of Argonne National Laboratory (Argonne) under Contracts No. DE-AC02-06CH11357 and DE-FG02-05ER25694 with the U.S. Department of Energy. The U.S. Government retains for itself, and others acting on its behalf, a paid-up, nonexclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, by or on behalf of the Government. The NCAR is sponsored by the National Science Foundation.

Funders	Funder number
U.S. Department of Energy
National Science Foundation
U.S. Department of Energy
Office of Science
Advanced Scientific Computing Research	DE-AC02-06CH11357
Argonne National Laboratory

Access to Document

10.1007/978-3-319-17353-5_7

Cite this

Balaprakash, P., Alexeev, Y., Mickelson, S. A., Leyffer, S., Jacob, R., & Craig, A. (2015). Machine-learning-based load balancing for community ice code component in CESM. In O. Marques, M. Dayde, & K. Nakajima (Eds.), High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers (pp. 79-91). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8969). Springer Verlag. https://doi.org/10.1007/978-3-319-17353-5_7

Balaprakash, Prasanna ; Alexeev, Yuri ; Mickelson, Sheri A. et al. / Machine-learning-based load balancing for community ice code component in CESM. High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. editor / Osni Marques ; Michel Dayde ; Kengo Nakajima. Springer Verlag, 2015. pp. 79-91 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{71be6178d70949d299e1cfdfd24fecba,

title = "Machine-learning-based load balancing for community ice code component in CESM",

abstract = "Load balancing scientific codes on massively parallel architectures is becoming an increasingly challenging task. In this paper, we focus on the Community Earth System Model, a widely used climate modeling code. It comprises six components each of which exhibits different scalability patterns. Previously, an analytical performance model has been used to find optimal load-balancing parameter configurations for each component. Nevertheless, for the Community Ice Code component, the analytical performance model is too restrictive to capture its scalability patterns. We therefore developed machine-learning-based load-balancing algorithm. It involves fitting a surrogate model to a small number of load-balancing configurations and their corresponding runtimes. This model is then used to find high-quality parameter configurations. Compared with the current practice of expert-knowledge-based enumeration over feasible configurations, the machine-learning-based load-balancing algorithm requires six times fewer evaluations to find the optimal configuration.",

author = "Prasanna Balaprakash and Yuri Alexeev and Mickelson, {Sheri A.} and Sven Leyffer and Robert Jacob and Anthony Craig",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014 ; Conference date: 30-06-2014 Through 03-07-2014",

year = "2015",

doi = "10.1007/978-3-319-17353-5_7",

language = "English",

isbn = "9783319173528",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "79--91",

editor = "Osni Marques and Michel Dayde and Kengo Nakajima",

booktitle = "High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers",

}

Balaprakash, P, Alexeev, Y, Mickelson, SA, Leyffer, S, Jacob, R & Craig, A 2015, Machine-learning-based load balancing for community ice code component in CESM. in O Marques, M Dayde & K Nakajima (eds), High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8969, Springer Verlag, pp. 79-91, 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014, Eugene, United States, 06/30/14. https://doi.org/10.1007/978-3-319-17353-5_7

Machine-learning-based load balancing for community ice code component in CESM. / Balaprakash, Prasanna; Alexeev, Yuri; Mickelson, Sheri A. et al.
High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. ed. / Osni Marques; Michel Dayde; Kengo Nakajima. Springer Verlag, 2015. p. 79-91 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8969).

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

TY - GEN

T1 - Machine-learning-based load balancing for community ice code component in CESM

AU - Balaprakash, Prasanna

AU - Alexeev, Yuri

AU - Mickelson, Sheri A.

AU - Leyffer, Sven

AU - Jacob, Robert

AU - Craig, Anthony

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2015.

PY - 2015

Y1 - 2015

N2 - Load balancing scientific codes on massively parallel architectures is becoming an increasingly challenging task. In this paper, we focus on the Community Earth System Model, a widely used climate modeling code. It comprises six components each of which exhibits different scalability patterns. Previously, an analytical performance model has been used to find optimal load-balancing parameter configurations for each component. Nevertheless, for the Community Ice Code component, the analytical performance model is too restrictive to capture its scalability patterns. We therefore developed machine-learning-based load-balancing algorithm. It involves fitting a surrogate model to a small number of load-balancing configurations and their corresponding runtimes. This model is then used to find high-quality parameter configurations. Compared with the current practice of expert-knowledge-based enumeration over feasible configurations, the machine-learning-based load-balancing algorithm requires six times fewer evaluations to find the optimal configuration.

AB - Load balancing scientific codes on massively parallel architectures is becoming an increasingly challenging task. In this paper, we focus on the Community Earth System Model, a widely used climate modeling code. It comprises six components each of which exhibits different scalability patterns. Previously, an analytical performance model has been used to find optimal load-balancing parameter configurations for each component. Nevertheless, for the Community Ice Code component, the analytical performance model is too restrictive to capture its scalability patterns. We therefore developed machine-learning-based load-balancing algorithm. It involves fitting a surrogate model to a small number of load-balancing configurations and their corresponding runtimes. This model is then used to find high-quality parameter configurations. Compared with the current practice of expert-knowledge-based enumeration over feasible configurations, the machine-learning-based load-balancing algorithm requires six times fewer evaluations to find the optimal configuration.

UR - http://www.scopus.com/inward/record.url?scp=84942593073&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-17353-5_7

DO - 10.1007/978-3-319-17353-5_7

M3 - Conference contribution

AN - SCOPUS:84942593073

SN - 9783319173528

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 79

EP - 91

BT - High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers

A2 - Marques, Osni

A2 - Dayde, Michel

A2 - Nakajima, Kengo

PB - Springer Verlag

T2 - 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014

Y2 - 30 June 2014 through 3 July 2014

ER -

Balaprakash P, Alexeev Y, Mickelson SA, Leyffer S, Jacob R, Craig A. Machine-learning-based load balancing for community ice code component in CESM. In Marques O, Dayde M, Nakajima K, editors, High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. Springer Verlag. 2015. p. 79-91. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-17353-5_7

Machine-learning-based load balancing for community ice code component in CESM

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

Cite this