Design optimization of the Grid data analysis workflow in CMS
Hassen Riahi · CERN Document Server (European Organization for Nuclear Research) · 2012
The chain of the typical CMS analysis workflow execution starts once configured and submitted by the physicist and ends when the outputs become available in the physicist storage area. During the execution, the workflow interacts with CMS computing infrastructure and services. These workflows have shown several bottlenecks through the years introducing delays in the execution of the end-user analysis and consequently differing the availability of physics results to the collaboration and squandering often the distributed resources. This thesis focuses on the study aiming to optimize the design of data analysis workflow of CMS, executing over LHC distributed computing infrastructure. Automation tool and an asynchronous stage-out system have been designed and implemented to reach this goal. The performance of the new tools in production environment are also shown.