A detailed analysis of communication load balance on BlueGene supercomputer

Yongzhi Chen; Yuefan Deng

doi:10.1016/j.cpc.2009.02.006

A detailed analysis of communication load balance on BlueGene supercomputer

Yongzhi Chen, Yuefan Deng

Mathematics

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

Abstract

We have developed several models that abstract applications as demand matrices and computer networks as supply matrices for minimizing the time to complete an application on a given computer by mapping. The mappings generated by our models improve the BlueGene/L performance of several engineering applications by up to 22% compared to the standard approach. Such improvements result mainly from balancing the communication loads among the network links.

Original language	English (US)
Pages (from-to)	1251-1258
Number of pages	8
Journal	Computer Physics Communications
Volume	180
Issue number	8
DOIs	https://doi.org/10.1016/j.cpc.2009.02.006
State	Published - Aug 2009

Keywords

BlueGene/L
Communication load balance
Performance analysis
Static mapping

ASJC Scopus subject areas

Hardware and Architecture
General Physics and Astronomy

Access to Document

10.1016/j.cpc.2009.02.006

Cite this

@article{e035e07bf67f4f8abb228d0a8609975a,

title = "A detailed analysis of communication load balance on BlueGene supercomputer",

abstract = "We have developed several models that abstract applications as demand matrices and computer networks as supply matrices for minimizing the time to complete an application on a given computer by mapping. The mappings generated by our models improve the BlueGene/L performance of several engineering applications by up to 22% compared to the standard approach. Such improvements result mainly from balancing the communication loads among the network links.",

keywords = "BlueGene/L, Communication load balance, Performance analysis, Static mapping",

author = "Yongzhi Chen and Yuefan Deng",

note = "Funding Information: This research utilized resources at the New York Center for Computational Sciences at Stony Brook University/Brookhaven National Laboratory which is supported by the U.S. Department of Energy under Contract No. DE-AC02-98CH10886 and by the State of New York. We gratefully acknowledge use of “BGL”, a 1024-node IBM Blue Gene/L system operated by the Argonne Leadership Computing Facility at Argonne National Laboratory.",

year = "2009",

month = aug,

doi = "10.1016/j.cpc.2009.02.006",

language = "English (US)",

volume = "180",

pages = "1251--1258",

journal = "Computer Physics Communications",

issn = "0010-4655",

publisher = "Elsevier",

number = "8",

}

TY - JOUR

T1 - A detailed analysis of communication load balance on BlueGene supercomputer

AU - Chen, Yongzhi

AU - Deng, Yuefan

N1 - Funding Information: This research utilized resources at the New York Center for Computational Sciences at Stony Brook University/Brookhaven National Laboratory which is supported by the U.S. Department of Energy under Contract No. DE-AC02-98CH10886 and by the State of New York. We gratefully acknowledge use of “BGL”, a 1024-node IBM Blue Gene/L system operated by the Argonne Leadership Computing Facility at Argonne National Laboratory.

PY - 2009/8

Y1 - 2009/8

N2 - We have developed several models that abstract applications as demand matrices and computer networks as supply matrices for minimizing the time to complete an application on a given computer by mapping. The mappings generated by our models improve the BlueGene/L performance of several engineering applications by up to 22% compared to the standard approach. Such improvements result mainly from balancing the communication loads among the network links.

AB - We have developed several models that abstract applications as demand matrices and computer networks as supply matrices for minimizing the time to complete an application on a given computer by mapping. The mappings generated by our models improve the BlueGene/L performance of several engineering applications by up to 22% compared to the standard approach. Such improvements result mainly from balancing the communication loads among the network links.

KW - BlueGene/L

KW - Communication load balance

KW - Performance analysis

KW - Static mapping

UR - http://www.scopus.com/inward/record.url?scp=67650866273&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=67650866273&partnerID=8YFLogxK

U2 - 10.1016/j.cpc.2009.02.006

DO - 10.1016/j.cpc.2009.02.006

M3 - Article

AN - SCOPUS:67650866273

SN - 0010-4655

VL - 180

SP - 1251

EP - 1258

JO - Computer Physics Communications

JF - Computer Physics Communications

IS - 8

ER -

A detailed analysis of communication load balance on BlueGene supercomputer

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this