Non-data-communication overheads in MPI: Analysis on blue Gene/P

Pavan Balaji; Anthony Chan; William Gropp; Rajeev Thakur; Ewing Lusk

doi:10.1007/978-3-540-87475-1_9

Non-data-communication overheads in MPI: Analysis on blue Gene/P

Pavan Balaji, Anthony Chan, William Gropp, Rajeev Thakur, Ewing Lusk

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

Abstract

Modern HEC systems, such as Blue Gene/P, rely on achieving high-performance by using the parallelism of a massive number of low-frequency/low-power processing cores. This means that the local pre- and post-communication processing required by the MPI stack might not be very fast, owing to the slow processing cores. Similarly, small amounts of serialization within the MPI stack that were acceptable on small/medium systems can be brutal on massively parallel systems. In this paper, we study different non-data-communication overheads within the MPI implementation on the IBM Blue Gene/P system.

Original language	English (US)
Title of host publication	Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings
Pages	13-22
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-540-87475-1_9
State	Published - 2008
Event	15th European PVM/MPI Users' Group Meeting, EuroPVM/MPI 2008 - Dublin, Ireland Duration: Sep 7 2008 → Sep 10 2008

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	5205 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	15th European PVM/MPI Users' Group Meeting, EuroPVM/MPI 2008
Country/Territory	Ireland
City	Dublin
Period	9/7/08 → 9/10/08

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Online availability

10.1007/978-3-540-87475-1_9

Library availability

Discover UIUC Full Text

Cite this

Balaji, P., Chan, A., Gropp, W., Thakur, R., & Lusk, E. (2008). Non-data-communication overheads in MPI: Analysis on blue Gene/P. In Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings (pp. 13-22). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5205 LNCS). https://doi.org/10.1007/978-3-540-87475-1_9

Non-data-communication overheads in MPI: Analysis on blue Gene/P. / Balaji, Pavan; Chan, Anthony; Gropp, William et al.
Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings. 2008. p. 13-22 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5205 LNCS).

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

Balaji, P, Chan, A, Gropp, W, Thakur, R & Lusk, E 2008, Non-data-communication overheads in MPI: Analysis on blue Gene/P. in Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5205 LNCS, pp. 13-22, 15th European PVM/MPI Users' Group Meeting, EuroPVM/MPI 2008, Dublin, Ireland, 9/7/08. https://doi.org/10.1007/978-3-540-87475-1_9

Balaji P, Chan A, Gropp W, Thakur R, Lusk E. Non-data-communication overheads in MPI: Analysis on blue Gene/P. In Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings. 2008. p. 13-22. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-540-87475-1_9

Balaji, Pavan ; Chan, Anthony ; Gropp, William et al. / Non-data-communication overheads in MPI : Analysis on blue Gene/P. Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings. 2008. pp. 13-22 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{d592cfbfcc0a439b87657ea086021884,

title = "Non-data-communication overheads in MPI: Analysis on blue Gene/P",

abstract = "Modern HEC systems, such as Blue Gene/P, rely on achieving high-performance by using the parallelism of a massive number of low-frequency/low-power processing cores. This means that the local pre- and post-communication processing required by the MPI stack might not be very fast, owing to the slow processing cores. Similarly, small amounts of serialization within the MPI stack that were acceptable on small/medium systems can be brutal on massively parallel systems. In this paper, we study different non-data-communication overheads within the MPI implementation on the IBM Blue Gene/P system.",

author = "Pavan Balaji and Anthony Chan and William Gropp and Rajeev Thakur and Ewing Lusk",

year = "2008",

doi = "10.1007/978-3-540-87475-1_9",

language = "English (US)",

isbn = "3540874747",

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

pages = "13--22",

booktitle = "Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings",

}

TY - GEN

T1 - Non-data-communication overheads in MPI

T2 - 15th European PVM/MPI Users' Group Meeting, EuroPVM/MPI 2008

AU - Balaji, Pavan

AU - Chan, Anthony

AU - Gropp, William

AU - Thakur, Rajeev

AU - Lusk, Ewing

PY - 2008

Y1 - 2008

N2 - Modern HEC systems, such as Blue Gene/P, rely on achieving high-performance by using the parallelism of a massive number of low-frequency/low-power processing cores. This means that the local pre- and post-communication processing required by the MPI stack might not be very fast, owing to the slow processing cores. Similarly, small amounts of serialization within the MPI stack that were acceptable on small/medium systems can be brutal on massively parallel systems. In this paper, we study different non-data-communication overheads within the MPI implementation on the IBM Blue Gene/P system.

AB - Modern HEC systems, such as Blue Gene/P, rely on achieving high-performance by using the parallelism of a massive number of low-frequency/low-power processing cores. This means that the local pre- and post-communication processing required by the MPI stack might not be very fast, owing to the slow processing cores. Similarly, small amounts of serialization within the MPI stack that were acceptable on small/medium systems can be brutal on massively parallel systems. In this paper, we study different non-data-communication overheads within the MPI implementation on the IBM Blue Gene/P system.

UR - http://www.scopus.com/inward/record.url?scp=56449088365&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56449088365&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-87475-1_9

DO - 10.1007/978-3-540-87475-1_9

M3 - Conference contribution

AN - SCOPUS:56449088365

SN - 3540874747

SN - 9783540874744

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

SP - 13

EP - 22

BT - Recent Advances in Parallel Virtual Machine and Message Passing Interface - 15th European PVM/MPI Users' Group Meeting, Proceedings

Y2 - 7 September 2008 through 10 September 2008

ER -

Non-data-communication overheads in MPI: Analysis on blue Gene/P

Abstract

Publication series

Other

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this