IGGI, A computing Framework for large Scale Parametric Simulations, Application to uncertainty analysis with TOUGHREACT

Abstract : IGGI stands for infrastructure for grids, cluster and intranet. This research project partially funded by the French government is aiming at developing technologies allowing the access and the gathering of the whole computing resources spread over the intranet of a company. This could include dedicated computing power or personal computers. The project is collaboration between BRGM, INRIA and Mandriva. Scientific computing has evolved for last few years and commodity-based components can run a wide range of applications. In this study, our focus will be on the software environment required to carry out large scale parametric simulations, but we will also report on experience of the deployment of such an infrastructure in the day to day life of an intranet. The IGGI software suite is mainly based on two components: ComputeModeTM which smoothly aggregates idle user machine to a virtual computing cluster. This is done through a transparent switch of the PC to a secondary, protected mode from which it boots from the ComputeMode server taking advantage of the PXE protocol. The second IGGI component is CIGRI which scheduled and executed computing tasks on the idle nodes of clusters. Special attention has been paid on the end-users ability to perform easily parametric simulations. Transparent access to the batch-scheduler, checkpoint and migration of the application will be exposed for the test-case of the analysis of uncertainty with TOUGHREACT : "Uncertainty in predictions of transfer model response to a thermal and alkaline perturbation in clay" . The calculations reported were carried out using idle computing capacity on personal computers inside the BRGM. This new architecture is particularly well suited to explore the wide range
Type de document :
Communication dans un congrès
TOUGH Symposium 2006, Lawrence Berkeley National Laboratory, May 2006, Berkeley, United States. 2006
Liste complète des métadonnées

https://hal.inria.fr/hal-00691559
Contributeur : Ist Rennes <>
Soumis le : jeudi 26 avril 2012 - 15:28:10
Dernière modification le : jeudi 11 octobre 2018 - 08:48:02

Identifiants

  • HAL Id : hal-00691559, version 1

Collections

Citation

Fabrice Dupros, Faïza Boulahya, Jacques Vairon, Pierre Lombard, Nicolas Capit, et al.. IGGI, A computing Framework for large Scale Parametric Simulations, Application to uncertainty analysis with TOUGHREACT. TOUGH Symposium 2006, Lawrence Berkeley National Laboratory, May 2006, Berkeley, United States. 2006. 〈hal-00691559〉

Partager

Métriques

Consultations de la notice

248