Experience with the gLite Workload Management System in ATLAS Monte Carlo production on LCG

The ATLAS experiment has been running continuous simulated events production since more than two years. A considerable fraction of the jobs is daily submitted and handled via the gLite Workload Management System, which overcomes several limitationsof the previous LCG Resource Broker. The gLite WMS has been tested very intensively for the LHC experiments use cases for more than six months, both in terms of performance and reliability. The tests were carried out by the LCG Experiment Integration Support team (in close contact with the experiments) together with the EGEE integration and certification team and the gLite middleware developers. A pragmatic iterative and interactive approach allowed a very quick rollout of fixes and their rapid deployment, together with new functionalities, for the ATLAS production activities. The same approach is being adopted for other middleware components like the gLite and CREAM Computing Elements. In this contribution we will summarize the learning from the gLite WMS testing activity, pointing out the most important achievements and the open issues. In addition, we will present the current situation of the ATLAS simulated event production activity on the EGEE infrastructure based on the gLite WMS, showing the main improvements and benefits from the new middleware. Finally, some preliminary results on the new flavors of Computing Elements usage will be shown, trying to identify possible advantag es not only in terms of robustness and performance, but also functionality for the experiment activities.