Dynamic load balancing is an important step conditioning the performance of parallel adaptive codes whose load evolution is difficult to predict. Most of the studies which answer this problem perform well, but are limited to an initially fixed number of processors which is not modified at runtime. These approaches can be very inefficient, especially in terms of resource consumption. In this paper, we present a new graph repartitioning algorithm which accepts to dynamically change the number of processors, assuming the load is already balanced. Our algorithm minimizes both data communication and data migration overheads, while maintaining the computational load balanced. This algorithm is based on a theoretical result, that constructs optimal communication patterns with both a minimum migration volume and a minimum number of communications. An experimental study which compares our work against stateof-the-art approaches is presented.
To cite this version:Clément Vuchener, Aurélien Esnard. Équilibrage dynamique avec nombre variable de processeurs par une méthode de repartitionnement de graphe. Revue des Sciences et Technologies de l'Information -Série TSI : Technique et Science Informatiques, Lavoisier, 2012Lavoisier, , 31 (8-9-10/2012 RÉSUMÉ. L'équilibrage dynamique de charge est une étape cruciale qui conditionne la performance des codes adaptatifs dont l'évolution de la charge est difficilement prévisible. Néan-moins, l'ensemble des travaux dans ce domaine se limitent -à notre connaissance -au cas où le nombre de processeurs est fixé initialement et n'est pas remis en cause lors de l'équilibrage. Cela peut s'avérer particulièrement inefficace, notamment du point de vue de la consommation des ressources. Nous proposons dans cet article deux nouveaux algorithmes de repartitionnement de graphe permettant de faire varier le nombre de processeurs, en supposant que la charge du graphe est déjà équilibrée. Ces algorithmes optimisent conjointement la coupe et la migration des données en s'appuyant sur un modèle de partitionnement de graphe à sommets fixes. Des résultats expérimentaux valident nos travaux en les comparant à d'autres approches.ABSTRACT. Dynamic load balancing is an important step affecting the performance of adaptive codes whose load evolution is difficult to predict. Nevertheless, as far as we know, other studies are limited to an initially fixed number of processors which is not modified during the balancing phase. It can be very inefficient, more particularly in terms of resource consumption. In this paper, we present two new graph repartitioning algorithms which permit a variable number of processors, assuming the load is already balanced. These algorithms optimize both edge cut and data migration using graph partitioning with fixed vertices. Experimental results validate our work comparing it with other approaches. MOTS-CLÉS : calcul haute-performance, équilibrage dynamique de charge, partitionnement de graphe.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2025 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.