Load balancing for regular meshes on SMPs with MPI

Vivek Kale; William Gropp

doi:10.1007/978-3-642-15646-5_24

Load balancing for regular meshes on SMPs with MPI

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

Abstract

Domain decomposition for regular meshes on parallel computers has traditionally been performed by attempting to exactly partition the work among the available processors (now cores). However, these strategies often do not consider the inherent system noise which can hinder MPI application scalability to emerging peta-scale machines with 10000+ nodes. In this work, we suggest a solution that uses a tunable hybrid static/dynamic scheduling strategy that can be incorporated into current MPI implementations of mesh codes. By applying this strategy to a 3D jacobi algorithm, we achieve performance gains of at least 16% for 64 SMP nodes.

Original language	English (US)
Title of host publication	Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings
Pages	229-238
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-15646-5_24
State	Published - 2010
Event	17th European MPI Users' Group Meeting, EuroMPI 2010 - Stuttgart, Germany Duration: Sep 12 2010 → Sep 15 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	6305 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	17th European MPI Users' Group Meeting, EuroMPI 2010
Country/Territory	Germany
City	Stuttgart
Period	9/12/10 → 9/15/10

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Online availability

10.1007/978-3-642-15646-5_24

Library availability

Discover UIUC Full Text

Cite this

Kale, V., & Gropp, W. (2010). Load balancing for regular meshes on SMPs with MPI. In Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings (pp. 229-238). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6305 LNCS). https://doi.org/10.1007/978-3-642-15646-5_24

Load balancing for regular meshes on SMPs with MPI. / Kale, Vivek; Gropp, William.
Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. 2010. p. 229-238 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6305 LNCS).

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

Kale, V & Gropp, W 2010, Load balancing for regular meshes on SMPs with MPI. in Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6305 LNCS, pp. 229-238, 17th European MPI Users' Group Meeting, EuroMPI 2010, Stuttgart, Germany, 9/12/10. https://doi.org/10.1007/978-3-642-15646-5_24

Kale V, Gropp W. Load balancing for regular meshes on SMPs with MPI. In Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. 2010. p. 229-238. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-15646-5_24

@inproceedings{4f7c8ab928844afe9197bc96f227e6ba,

title = "Load balancing for regular meshes on SMPs with MPI",

abstract = "Domain decomposition for regular meshes on parallel computers has traditionally been performed by attempting to exactly partition the work among the available processors (now cores). However, these strategies often do not consider the inherent system noise which can hinder MPI application scalability to emerging peta-scale machines with 10000+ nodes. In this work, we suggest a solution that uses a tunable hybrid static/dynamic scheduling strategy that can be incorporated into current MPI implementations of mesh codes. By applying this strategy to a 3D jacobi algorithm, we achieve performance gains of at least 16% for 64 SMP nodes.",

author = "Vivek Kale and William Gropp",

year = "2010",

doi = "10.1007/978-3-642-15646-5_24",

language = "English (US)",

isbn = "3642156452",

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

pages = "229--238",

booktitle = "Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings",

}

TY - GEN

T1 - Load balancing for regular meshes on SMPs with MPI

AU - Kale, Vivek

AU - Gropp, William

PY - 2010

Y1 - 2010

N2 - Domain decomposition for regular meshes on parallel computers has traditionally been performed by attempting to exactly partition the work among the available processors (now cores). However, these strategies often do not consider the inherent system noise which can hinder MPI application scalability to emerging peta-scale machines with 10000+ nodes. In this work, we suggest a solution that uses a tunable hybrid static/dynamic scheduling strategy that can be incorporated into current MPI implementations of mesh codes. By applying this strategy to a 3D jacobi algorithm, we achieve performance gains of at least 16% for 64 SMP nodes.

AB - Domain decomposition for regular meshes on parallel computers has traditionally been performed by attempting to exactly partition the work among the available processors (now cores). However, these strategies often do not consider the inherent system noise which can hinder MPI application scalability to emerging peta-scale machines with 10000+ nodes. In this work, we suggest a solution that uses a tunable hybrid static/dynamic scheduling strategy that can be incorporated into current MPI implementations of mesh codes. By applying this strategy to a 3D jacobi algorithm, we achieve performance gains of at least 16% for 64 SMP nodes.

UR - http://www.scopus.com/inward/record.url?scp=78149277691&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149277691&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15646-5_24

DO - 10.1007/978-3-642-15646-5_24

M3 - Conference contribution

AN - SCOPUS:78149277691

SN - 3642156452

SN - 9783642156458

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

SP - 229

EP - 238

BT - Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings

T2 - 17th European MPI Users' Group Meeting, EuroMPI 2010

Y2 - 12 September 2010 through 15 September 2010

ER -

Load balancing for regular meshes on SMPs with MPI

Abstract

Publication series

Other

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this