The usage of the gLite Workload Management System by the LHC experiments

The LHC experiments will study the physics of p-p interactions at a centre-of-mass energy of 14 TeV using the LHC accelerator at CERN. The primary purpose of their research is the discovery of the Higgs boson and of new physics at the TeV scale. The LHC experiments are undoubtedly the most demanding communities for the EGEE infrastructure. They have thousands of collaborators who expect to run their physics analyses on the data that will be collected starting from late 2007. It is expected that each experiment will need to run hundreds of thousands of jobs per day, including: event reconstruction from raw data, analysis on reconstructed data, and Monte Carlo simulation. The management of such amounts of jobs is an extremely complicated task and the gLite WMS is being considered as a candidate to perform it. This report describes the experience of the ATLAS and the CMS experiments using the gLite Workload Management System as a tool to submit both simulation and analysis jobs. In fact, this experience has led to a significant improvement of the WMS performance and reliability due to a close interaction with the gLite developers. Many limitations have been hit, and solved or alleviated. An evaluation of the current status of the WMS and how it matches with the experiment's expectations will be given. The most severe limitations found during this work relate to the ability of the WMS of working reliably and unattended for long periods of time. Though a lot of progress has been made, some issues remain. There are proposals to solve them, and they will be discussed.