CRAB: Distributed analysis tool for CMS

Abstract

CMS has a distributed computing model, based on a hierarchy of tiered regional computing centers and adopts a data driven model for the end user analysis. This model foresees that jobs are submitted to the analysis resources where data are hosted. The increasing complexity of the whole computing infrastructure makes the simple analysis work flow more and more complicated for the end user. CMS has developed and deployed a dedicated tool named CRAB (CMS Remote Analysis Builder) in order to guarantee the physicists an efficient access to the distributed data whilst hiding the underlying complexity. This tool is used by CMS to enable the running of physics analysis jobs in a transparent manner over data distributed across sites. It factorizes out the interaction with the underlying batch farms, grid infrastructure and CMS data management tools, allowing the user to deal only with a simple and intuitive interface. We present the CRAB architecture, as well as the current status and lessons learnt in deploying this tool for use by the CMS collaboration. We also present the future development of the CRAB system.