Collective communication costs analysis over Gigabit Ethernet and InfiniBand

Hyacinthe Nzigou Mamadou, Takeshi Nanri, Kazuaki Murakami

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

Users of parallel machines need to have a good grasp for how different communication patterns and styles affect the performance of message-passing applications. MPI Collective communications involve multiple processors, and their performance prediction is a tricky task to perform. In order to evaluate the performance of collective communications, we attempt to extend LogGP and P-LogP standard point-to-point models. Our objective is to compare these models with the empirical data, and identify the most suitable for performance characterization of collective communications. The models proposed are related with the implemented algorithms in MPICH. The experimental results performed on clusters of 16 and 64 processors connected by Infiniband and Gigabit Ethernet networks respectively, show the same trend. For any collective operation, given a number of processors and a range of message sizes, there is at least one model that predicts the performance precisely.

Original languageEnglish
Title of host publicationHigh Performance Computing - HiPC 2006 - 13th International Conference Proceedings
Pages547-559
Number of pages13
DOIs
Publication statusPublished - Dec 1 2006
Event13th International Conference on High Performance Computing, HiPC 2006 - Bangalore, India
Duration: Dec 18 2006Dec 21 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4297 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other13th International Conference on High Performance Computing, HiPC 2006
CountryIndia
CityBangalore
Period12/18/0612/21/06

Fingerprint

Collective Communication
InfiniBand
Cost Analysis
Ethernet
Communication Cost
Communication
Costs
Parallel Machines
Message passing
Performance Prediction
Message Passing
Model
Predict
Evaluate
Experimental Results
Range of data

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Mamadou, H. N., Nanri, T., & Murakami, K. (2006). Collective communication costs analysis over Gigabit Ethernet and InfiniBand. In High Performance Computing - HiPC 2006 - 13th International Conference Proceedings (pp. 547-559). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4297 LNCS). https://doi.org/10.1007/11945918_52

Collective communication costs analysis over Gigabit Ethernet and InfiniBand. / Mamadou, Hyacinthe Nzigou; Nanri, Takeshi; Murakami, Kazuaki.

High Performance Computing - HiPC 2006 - 13th International Conference Proceedings. 2006. p. 547-559 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4297 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Mamadou, HN, Nanri, T & Murakami, K 2006, Collective communication costs analysis over Gigabit Ethernet and InfiniBand. in High Performance Computing - HiPC 2006 - 13th International Conference Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4297 LNCS, pp. 547-559, 13th International Conference on High Performance Computing, HiPC 2006, Bangalore, India, 12/18/06. https://doi.org/10.1007/11945918_52
Mamadou HN, Nanri T, Murakami K. Collective communication costs analysis over Gigabit Ethernet and InfiniBand. In High Performance Computing - HiPC 2006 - 13th International Conference Proceedings. 2006. p. 547-559. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/11945918_52
Mamadou, Hyacinthe Nzigou ; Nanri, Takeshi ; Murakami, Kazuaki. / Collective communication costs analysis over Gigabit Ethernet and InfiniBand. High Performance Computing - HiPC 2006 - 13th International Conference Proceedings. 2006. pp. 547-559 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{473b3d91cbe74ac5a7467d3a89430648,
title = "Collective communication costs analysis over Gigabit Ethernet and InfiniBand",
abstract = "Users of parallel machines need to have a good grasp for how different communication patterns and styles affect the performance of message-passing applications. MPI Collective communications involve multiple processors, and their performance prediction is a tricky task to perform. In order to evaluate the performance of collective communications, we attempt to extend LogGP and P-LogP standard point-to-point models. Our objective is to compare these models with the empirical data, and identify the most suitable for performance characterization of collective communications. The models proposed are related with the implemented algorithms in MPICH. The experimental results performed on clusters of 16 and 64 processors connected by Infiniband and Gigabit Ethernet networks respectively, show the same trend. For any collective operation, given a number of processors and a range of message sizes, there is at least one model that predicts the performance precisely.",
author = "Mamadou, {Hyacinthe Nzigou} and Takeshi Nanri and Kazuaki Murakami",
year = "2006",
month = "12",
day = "1",
doi = "10.1007/11945918_52",
language = "English",
isbn = "354068039X",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "547--559",
booktitle = "High Performance Computing - HiPC 2006 - 13th International Conference Proceedings",

}

TY - GEN

T1 - Collective communication costs analysis over Gigabit Ethernet and InfiniBand

AU - Mamadou, Hyacinthe Nzigou

AU - Nanri, Takeshi

AU - Murakami, Kazuaki

PY - 2006/12/1

Y1 - 2006/12/1

N2 - Users of parallel machines need to have a good grasp for how different communication patterns and styles affect the performance of message-passing applications. MPI Collective communications involve multiple processors, and their performance prediction is a tricky task to perform. In order to evaluate the performance of collective communications, we attempt to extend LogGP and P-LogP standard point-to-point models. Our objective is to compare these models with the empirical data, and identify the most suitable for performance characterization of collective communications. The models proposed are related with the implemented algorithms in MPICH. The experimental results performed on clusters of 16 and 64 processors connected by Infiniband and Gigabit Ethernet networks respectively, show the same trend. For any collective operation, given a number of processors and a range of message sizes, there is at least one model that predicts the performance precisely.

AB - Users of parallel machines need to have a good grasp for how different communication patterns and styles affect the performance of message-passing applications. MPI Collective communications involve multiple processors, and their performance prediction is a tricky task to perform. In order to evaluate the performance of collective communications, we attempt to extend LogGP and P-LogP standard point-to-point models. Our objective is to compare these models with the empirical data, and identify the most suitable for performance characterization of collective communications. The models proposed are related with the implemented algorithms in MPICH. The experimental results performed on clusters of 16 and 64 processors connected by Infiniband and Gigabit Ethernet networks respectively, show the same trend. For any collective operation, given a number of processors and a range of message sizes, there is at least one model that predicts the performance precisely.

UR - http://www.scopus.com/inward/record.url?scp=77049123230&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77049123230&partnerID=8YFLogxK

U2 - 10.1007/11945918_52

DO - 10.1007/11945918_52

M3 - Conference contribution

AN - SCOPUS:77049123230

SN - 354068039X

SN - 9783540680390

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

SP - 547

EP - 559

BT - High Performance Computing - HiPC 2006 - 13th International Conference Proceedings

ER -