Skip to Main content Skip to Navigation

Gestion dynamique des tâches dans les grappes, une approche à base de machines virtuelles

Fabien Hermenier 1, 2, 3
1 ASCOLA - Aspect and composition languages
LINA - Laboratoire d'Informatique de Nantes Atlantique, Département informatique - EMN, Inria Rennes – Bretagne Atlantique
Abstract : Resources Management Systems relying on a dynamic management of jobs can efficiently use resources in clusters. Indeed they provide mechanisms to manipulate online the state of the jobs and their assignment on the nodes. In practice, these scheduling strategies are hard to deploy on clusters as they can not necessarily handle the manipulation of the jobs and may have specific scheduling constraints to consider. In this thesis, we try to ease the development of resources management systems relying on a dynamic management of jobs. We based our environment on the use of virtual machines to execute the jobs in their legacy environments while providing the mechanisms to manipulate them in a non-intrusive way. Moreover, we propose an autonomous environment to continuously optimize the scheduling of jobs. Scheduling strategies are implemented using constraints programming which aims to model and solve combinatory problems. We validate our approach with the development of our prototype Entropy, which has been used to implement various scheduling strategies. The evaluation of these strategies show their capability to to solve present problems.
Complete list of metadata
Contributor : Fabien Hermenier Connect in order to contact the contributor
Submitted on : Tuesday, April 27, 2010 - 1:01:43 PM
Last modification on : Friday, October 23, 2020 - 4:39:37 PM
Long-term archiving on: : Tuesday, September 28, 2010 - 1:00:15 PM


  • HAL Id : tel-00476822, version 1


Fabien Hermenier. Gestion dynamique des tâches dans les grappes, une approche à base de machines virtuelles. Réseaux et télécommunications [cs.NI]. Université de Nantes, 2009. Français. ⟨tel-00476822⟩



Record views


Files downloads