Fast and Efficient Total Exchange on Two Clusters

Emmanuel Jeannot 1 Luiz Angelo Steffenel 1
1 ALGORILLE - Algorithms for the Grid
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : Total Exchange is one of the most important collective communication patterns for scientific applications. In this paper we propose an algorithm called ${\mathcal LG}$ for the total exchange redistribution problem between two clusters. In our approach we perform communications in two different phases, aiming to minimize the number of communication steps through the wide-area network. Therefore, we are able to reduce the number of messages exchanged through the backbone to only 2× max (n 1 ,n 2) against 2×n 1 ×n 2 messages with the traditional strategy (where n 1 and n 2 are the number of nodes of each clusters). Experimental results show that we reach over than 50% of performance improvement comparing to the traditional strategies.
Type de document :
Communication dans un congrès
13th International European Conference on Parallel and Distributed Computing - Euro-Par 2007, Aug 2008, Rennes, France. Springer Berlin / Heidelberg, 4641, pp.848-857, 2007, Lecture Notes in Computer Science. 〈10.1007/978-3-540-74466-5_91〉
Liste complète des métadonnées

Littérature citée [19 références]  Voir  Masquer  Télécharger
https://hal.inria.fr/inria-00177533
Contributeur : Emmanuel Jeannot <>
Soumis le : mardi 16 octobre 2007 - 11:29:12
Dernière modification le : mercredi 14 février 2018 - 16:54:02
Document(s) archivé(s) le : lundi 24 septembre 2012 - 13:15:15

Fichier

europar07.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

Collections

INRIA | GRID5000 | LORIA | UNIV-LORRAINE

Citation

Emmanuel Jeannot, Luiz Angelo Steffenel. Fast and Efficient Total Exchange on Two Clusters. 13th International European Conference on Parallel and Distributed Computing - Euro-Par 2007, Aug 2008, Rennes, France. Springer Berlin / Heidelberg, 4641, pp.848-857, 2007, Lecture Notes in Computer Science. 〈10.1007/978-3-540-74466-5_91〉. 〈inria-00177533〉

Exporter

BibTeX TEI DC DCterms EndNote

Partager

Métriques

Consultations de la notice

196

Téléchargements de fichiers

100

Contact


archive-ouverte@inria.fr