Assessing Contention Effects on MPI_Alltoall Communications

Luiz Angelo Steffenel 1 Maxime Martinasso 2 Denis Trystram 3
1 ALGORILLE - Algorithms for the Grid
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
2 MESCAL - Middleware efficiently scalable
ID-IMAG - Informatique et Distribution, Inria Grenoble - Rhône-Alpes
3 MOAIS - PrograMming and scheduling design fOr Applications in Interactive Simulation
ID-IMAG - Informatique et Distribution, Inria Grenoble - Rhône-Alpes, UJF - Université Joseph Fourier - Grenoble 1
Abstract : One of the most important collective communication patterns used in scientific applications is the complete exchange, also called All-to-All. Although efficient algorithms have been studied for specific networks, general solutions like those available in well-known MPI distributions (e.g. the MPI_Alltoall operation) are strongly influenced by the congestion of network resources. In this paper we present an integrated approach to model the performance of the All-to-All collective operation, which consists in identifying a contention signature that characterizes a given network environment, using it to augment a contention-free communication model. This approach, assessed by experimental results, allows an accurate prediction of the performance of the All-to-All operation over different network architectures with a small overhead.
Liste complète des métadonnées

Cited literature [18 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-00136204
Contributor : Luiz Angelo Steffenel <>
Submitted on : Monday, March 12, 2007 - 6:55:53 PM
Last modification on : Wednesday, March 13, 2019 - 3:02:02 PM
Document(s) archivé(s) le : Friday, September 21, 2012 - 12:50:32 PM

File

GPC2007-Steffenel.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00136204, version 1

Collections

Citation

Luiz Angelo Steffenel, Maxime Martinasso, Denis Trystram. Assessing Contention Effects on MPI_Alltoall Communications. 2nd International Conference on Grid and Pervasive Computing - GPC 2007, May 2007, France. pp.00. ⟨hal-00136204⟩

Share

Metrics

Record views

621

Files downloads

364