Mainz 2014 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 35: GRID-Computing

T 35.8: Talk

Monday, March 24, 2014, 18:30–18:45, P15

Statistical analysis of Grid activities in a WLCG Tier-2 site — •Gen Kawamura, Jordi Nadal, and Arnulf Quadt — II. Physikalisches Institut, Georg-August Universität Göttingen, Friedrich-Hund-Platz 1, 37077, Göttingen, Deutschland

The World-Wide LHC Computing Grid (WLCG) sites have handled massive data and job requests produced by the LHC and the ATLAS experiments. The dCache storage management system and the CREAMCE job management system have been widely used by the HEP community in the LHC experiments, and their massive grid activities have been recorded. However, such massive activities are being increasingly utilized to understand the actual use of the grid and timely monitoring. Thus we must use significant computational resources for analyzing such activities. We consider a data reduction technique of such data under the conditions that sites are operational and are in production for the LHC experiments. Thus, the log data containing statistical variables are not assumed to be Gaussian. Therefore in this case, it is necessary to consider bootstrap approximations based on statistics, since the distribution of data is unknown. A random data selection of the bootstrap method can be applied, thus reducing the cost of computations, and instantly obtaining up-to-date information. Using results from a large-scale distributed WLCG Tier-2 site, GoeGrid, we introduce a new procedure which presents a computationally efficient method of assessing quality among grid users, clients, grid components and individual networks.