Ghost Process: a Sound Basis to Implement Process Duplication, Migration and Checkpoint/Restart in Linux Clusters - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Rapport (Rapport De Recherche) Année : 2005

Ghost Process: a Sound Basis to Implement Process Duplication, Migration and Checkpoint/Restart in Linux Clusters

Résumé

Today, clusters are widely used to execute numerical applications. Mechanisms are needed to ease cluster use and to take advantage of the cluster distributed resources. Process management mechanisms are very useful in this respect. Process duplication is needed for parallel application deployment, dynamic load balancing relies on process migration and process checkpoint/restart is needed to tolerate node failures during the execution of long-running applications. A kernel level approach allows to efficiently implement all these mechanisms. However, they are very complex to implement and to maintain. Nevertheless they are all based on a common concept: process virtualization which provides a mean to extract the state of a process from the operating system executed on each cluster node. This paper presents a process virtualization mechanism called ghost process, implemented at kernel level which can be used by system programmers to easily implement various process management mechanisms. The ghost process mechanism has been implemented in Kerrighed single system image cluster operating system based on Linux. It has been used to efficiently and easily implement process duplication, migration and checkpoint/restart in Kerrighed.
Fichier principal
Vignette du fichier
RR-5476.pdf (371.04 Ko) Télécharger le fichier

Dates et versions

inria-00070531 , version 1 (19-05-2006)

Identifiants

  • HAL Id : inria-00070531 , version 1

Citer

Geoffroy Vallée, Renaud Lottiaux, David Margery, Christine Morin, Jean-Yves Berthou. Ghost Process: a Sound Basis to Implement Process Duplication, Migration and Checkpoint/Restart in Linux Clusters. [Research Report] RR-5476, INRIA. 2005, pp.16. ⟨inria-00070531⟩
394 Consultations
132 Téléchargements

Partager

Gmail Facebook X LinkedIn More