Analysis of topology-dependent MPI performance on gemini networks

Antonio J. Peña; Ralf G.Correa Carvalho; James Dinan; Pavan Balaji; Rajeev Thakur; William Gropp

doi:10.1145/2488551.2488564

Analysis of topology-dependent MPI performance on gemini networks

Antonio J. Peña, Ralf G.Correa Carvalho, James Dinan, Pavan Balaji, Rajeev Thakur, William Gropp

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

Abstract

Current HPC systems utilize a variety of interconnection networks, with varying features and communication characteristics. MPI normalizes these interconnects with a common interface used by most HPC applications. However, network properties can have a significant impact on application performance. We explore the impact of the interconnect on application performance on the Blue Waters supercomputer. Blue Waters uses a three-dimensional, Cray Gemini torus network, which provides twice the Y-dimension bandwidth in the X and Z dimensions. Through several benchmarks, including a halo-exchange example, we demonstrate that application-level mapping to the network topology yields significant performance improvements.

Original language	English (US)
Title of host publication	Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013
Publisher	Association for Computing Machinery
Pages	61-66
Number of pages	6
ISBN (Print)	9788461651337
DOIs	https://doi.org/10.1145/2488551.2488564
State	Published - 2013
Event	20th European MPI Users' Group Meeting, EuroMPI 2013 - Madrid, Spain Duration: Sep 15 2013 → Sep 18 2013

Publication series

Name	ACM International Conference Proceeding Series

Other

Other	20th European MPI Users' Group Meeting, EuroMPI 2013
Country/Territory	Spain
City	Madrid
Period	9/15/13 → 9/18/13

Keywords

Gemini
Interconnection network
MPI
Network topology

ASJC Scopus subject areas

Software
Human-Computer Interaction
Computer Vision and Pattern Recognition
Computer Networks and Communications

Online availability

10.1145/2488551.2488564

Library availability

Discover UIUC Full Text

Cite this

Peña, A. J., Carvalho, R. G. C., Dinan, J., Balaji, P., Thakur, R., & Gropp, W. (2013). Analysis of topology-dependent MPI performance on gemini networks. In Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013 (pp. 61-66). (ACM International Conference Proceeding Series). Association for Computing Machinery. https://doi.org/10.1145/2488551.2488564

Analysis of topology-dependent MPI performance on gemini networks. / Peña, Antonio J.; Carvalho, Ralf G.Correa; Dinan, James et al.

Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013. Association for Computing Machinery, 2013. p. 61-66 (ACM International Conference Proceeding Series).

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

Peña, AJ, Carvalho, RGC, Dinan, J, Balaji, P, Thakur, R & Gropp, W 2013, Analysis of topology-dependent MPI performance on gemini networks. in Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 61-66, 20th European MPI Users' Group Meeting, EuroMPI 2013, Madrid, Spain, 9/15/13. https://doi.org/10.1145/2488551.2488564

@inproceedings{5a312028a62f4198a63aa5c1f9554052,

title = "Analysis of topology-dependent MPI performance on gemini networks",

abstract = "Current HPC systems utilize a variety of interconnection networks, with varying features and communication characteristics. MPI normalizes these interconnects with a common interface used by most HPC applications. However, network properties can have a significant impact on application performance. We explore the impact of the interconnect on application performance on the Blue Waters supercomputer. Blue Waters uses a three-dimensional, Cray Gemini torus network, which provides twice the Y-dimension bandwidth in the X and Z dimensions. Through several benchmarks, including a halo-exchange example, we demonstrate that application-level mapping to the network topology yields significant performance improvements.",

keywords = "Gemini, Interconnection network, MPI, Network topology",

author = "Pe{\~n}a, {Antonio J.} and Carvalho, {Ralf G.Correa} and James Dinan and Pavan Balaji and Rajeev Thakur and William Gropp",

year = "2013",

doi = "10.1145/2488551.2488564",

language = "English (US)",

isbn = "9788461651337",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "61--66",

booktitle = "Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013",

note = "20th European MPI Users' Group Meeting, EuroMPI 2013 ; Conference date: 15-09-2013 Through 18-09-2013",

}

TY - GEN

T1 - Analysis of topology-dependent MPI performance on gemini networks

AU - Peña, Antonio J.

AU - Carvalho, Ralf G.Correa

AU - Dinan, James

AU - Balaji, Pavan

AU - Thakur, Rajeev

AU - Gropp, William

PY - 2013

Y1 - 2013

N2 - Current HPC systems utilize a variety of interconnection networks, with varying features and communication characteristics. MPI normalizes these interconnects with a common interface used by most HPC applications. However, network properties can have a significant impact on application performance. We explore the impact of the interconnect on application performance on the Blue Waters supercomputer. Blue Waters uses a three-dimensional, Cray Gemini torus network, which provides twice the Y-dimension bandwidth in the X and Z dimensions. Through several benchmarks, including a halo-exchange example, we demonstrate that application-level mapping to the network topology yields significant performance improvements.

AB - Current HPC systems utilize a variety of interconnection networks, with varying features and communication characteristics. MPI normalizes these interconnects with a common interface used by most HPC applications. However, network properties can have a significant impact on application performance. We explore the impact of the interconnect on application performance on the Blue Waters supercomputer. Blue Waters uses a three-dimensional, Cray Gemini torus network, which provides twice the Y-dimension bandwidth in the X and Z dimensions. Through several benchmarks, including a halo-exchange example, we demonstrate that application-level mapping to the network topology yields significant performance improvements.

KW - Gemini

KW - Interconnection network

KW - MPI

KW - Network topology

UR - http://www.scopus.com/inward/record.url?scp=84886259134&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84886259134&partnerID=8YFLogxK

U2 - 10.1145/2488551.2488564

DO - 10.1145/2488551.2488564

M3 - Conference contribution

AN - SCOPUS:84886259134

SN - 9788461651337

T3 - ACM International Conference Proceeding Series

SP - 61

EP - 66

BT - Proceedings of the 20th European MPI Users' Group Meeting, EuroMPI 2013

PB - Association for Computing Machinery

T2 - 20th European MPI Users' Group Meeting, EuroMPI 2013

Y2 - 15 September 2013 through 18 September 2013

ER -

Analysis of topology-dependent MPI performance on gemini networks

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this