Understanding performance of SMP clusters running MPI programs

Franck Cappello; Olivier Richard; Daniel Etiemble

doi:10.1016/S0167-739X(00)00054-6

Understanding performance of SMP clusters running MPI programs

Franck Cappello, Olivier Richard, Daniel Etiemble

Research output: Contribution to journal › Article › peer-review

Abstract

Clusters of multiprocessors (CLUMPs) have an hybrid memory model, with message passing between nodes and shared memory inside nodes. We examine the performance of Myrinet clusters of SMP PCs when using a single memory model (SMM) based on the MPICH-PM/CLUMP library of the RWCP, which can directly use the MPI programs written for a cluster of uniprocessors. The specificities of the communication patterns with the SMM approach are detailed. PC clusters with 2-way and 4-way nodes are considered and compared.

Original language	English (US)
Pages (from-to)	711-720
Number of pages	10
Journal	Future Generation Computer Systems
Volume	17
Issue number	6
DOIs	https://doi.org/10.1016/S0167-739X(00)00054-6
State	Published - Apr 2001
Externally published	Yes

Keywords

Clusters of multiprocessors
MPI programs
SMP clusters

ASJC Scopus subject areas

Software
Hardware and Architecture
Computer Networks and Communications

Online availability

10.1016/S0167-739X(00)00054-6

Library availability

Discover UIUC Full Text

Cite this

@article{d19cb2207ce845fd96245f82c168d718,

title = "Understanding performance of SMP clusters running MPI programs",

abstract = "Clusters of multiprocessors (CLUMPs) have an hybrid memory model, with message passing between nodes and shared memory inside nodes. We examine the performance of Myrinet clusters of SMP PCs when using a single memory model (SMM) based on the MPICH-PM/CLUMP library of the RWCP, which can directly use the MPI programs written for a cluster of uniprocessors. The specificities of the communication patterns with the SMM approach are detailed. PC clusters with 2-way and 4-way nodes are considered and compared.",

keywords = "Clusters of multiprocessors, MPI programs, SMP clusters",

author = "Franck Cappello and Olivier Richard and Daniel Etiemble",

year = "2001",

month = apr,

doi = "10.1016/S0167-739X(00)00054-6",

language = "English (US)",

volume = "17",

pages = "711--720",

journal = "Future Generation Computer Systems",

issn = "0167-739X",

publisher = "Elsevier B.V.",

number = "6",

}

TY - JOUR

T1 - Understanding performance of SMP clusters running MPI programs

AU - Cappello, Franck

AU - Richard, Olivier

AU - Etiemble, Daniel

PY - 2001/4

Y1 - 2001/4

N2 - Clusters of multiprocessors (CLUMPs) have an hybrid memory model, with message passing between nodes and shared memory inside nodes. We examine the performance of Myrinet clusters of SMP PCs when using a single memory model (SMM) based on the MPICH-PM/CLUMP library of the RWCP, which can directly use the MPI programs written for a cluster of uniprocessors. The specificities of the communication patterns with the SMM approach are detailed. PC clusters with 2-way and 4-way nodes are considered and compared.

AB - Clusters of multiprocessors (CLUMPs) have an hybrid memory model, with message passing between nodes and shared memory inside nodes. We examine the performance of Myrinet clusters of SMP PCs when using a single memory model (SMM) based on the MPICH-PM/CLUMP library of the RWCP, which can directly use the MPI programs written for a cluster of uniprocessors. The specificities of the communication patterns with the SMM approach are detailed. PC clusters with 2-way and 4-way nodes are considered and compared.

KW - Clusters of multiprocessors

KW - MPI programs

KW - SMP clusters

UR - http://www.scopus.com/inward/record.url?scp=0035312315&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0035312315&partnerID=8YFLogxK

U2 - 10.1016/S0167-739X(00)00054-6

DO - 10.1016/S0167-739X(00)00054-6

M3 - Article

AN - SCOPUS:0035312315

SN - 0167-739X

VL - 17

SP - 711

EP - 720

JO - Future Generation Computer Systems

JF - Future Generation Computer Systems

IS - 6

ER -

Understanding performance of SMP clusters running MPI programs

Abstract

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this