591 articles – 2447 Notices  [english version]
HAL : in2p3-00737583, version 1

Fiche détaillée  Récupérer au format
17th International Conference on Computing in High Energy and Nuclear Physics (CHEP09), Prague : Tchèque, République (2010)
DIRAC pilot framework and the DIRAC Workload Management System
A. Casajus, R. Graciani, S. Paterson, A. Tsaregorodtsev1
LHCb Collaboration(s)

DIRAC, the LHCb community Grid solution, has pioneered the use of pilot jobs in the Grid. Pilot Jobs provide a homogeneous interface to an heterogeneous set of computing resources. At the same time, Pilot Jobs allow to delay the scheduling decision to the last moment, thus taking into account the precise running conditions at the resource and last moment requests to the system. The DIRAC Workload Management System provides one single scheduling mechanism for jobs with very different profiles. To achieve an overall optimisation, it organizes pending jobs in task queues, both for individual users and production activities. Task queues are created with jobs having similar requirements. Following the VO policy a priority is assigned to each task queue. Pilot submission and subsequent job matching are based on these priorities following a statistical approach.
1 :  CPPM - Centre de Physique des Particules de Marseille
Physique/Physique/Instrumentations et Détecteurs

Informatique/Calcul parallèle, distribué et partagé