Adaptive Dynamic Load Balancing in Heterogenous Multiple GPUs-CPUs Distributed Setting: Case Study of B&B Tree Search

Trong-Tuan Vu 1 Bilel Derbel 1, 2 Nouredine Melab 2
1 DOLPHIN - Parallel Cooperative Multi-criteria Optimization
LIFL - Laboratoire d'Informatique Fondamentale de Lille, Inria Lille - Nord Europe
Abstract : The emergence of new hybrid and heterogenous multi-GPU multi-CPU large scale platforms offers new opportunities and pauses new challenges when solving difficult optimization problems. This paper targets irregular tree search algorithms in which workload is unpredictable. We propose an adaptive distributed approach allowing to distribute the load dynamically at runtime while taking into account the computing abilities of either GPUs or CPUs. Using Branch-and-Bound and Flowshop as a case study, we deployed our approach using up to 20 GPUs jointly to up to 128 CPUs. Through extensive experiments in different system configurations, we report near optimal speedups, thus providing new insights into how to take full advantage of both GPUs and CPUs power in modern computing platforms.
Type de document :
Communication dans un congrès
7th International Learning and Intelligent OptimizatioN Conference (LION), Jan 2013, Catania, Italy. Lecture Notes in Computer Science, 2013
Liste complète des métadonnées

Littérature citée [13 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-00765199
Contributeur : Bilel Derbel <>
Soumis le : jeudi 7 mars 2013 - 13:02:28
Dernière modification le : jeudi 11 janvier 2018 - 06:22:13
Document(s) archivé(s) le : lundi 17 juin 2013 - 11:08:42

Fichier

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

Identifiants

  • HAL Id : hal-00765199, version 1

Citation

Trong-Tuan Vu, Bilel Derbel, Nouredine Melab. Adaptive Dynamic Load Balancing in Heterogenous Multiple GPUs-CPUs Distributed Setting: Case Study of B&B Tree Search. 7th International Learning and Intelligent OptimizatioN Conference (LION), Jan 2013, Catania, Italy. Lecture Notes in Computer Science, 2013. 〈hal-00765199〉

Partager

Métriques

Consultations de la notice

323

Téléchargements de fichiers

434