The CMS Computing Model with a Focus on the German T1 Activities

V.Buege / A. Scheurer
U Karlsruhe

Experiments at modern hadron colliders have to deal with data production
rates and event sizes which exceed those of former experiments by far. The
four experiments of the Large Hadron Collider (LHC) have decided to cope
with this challenge using grid technologies to ensure redundant data storage
and access to the data for the physicists all over the world. However,
besides the advantages of such a decentralised system, the administration of
these heterogeneous resources is quite challenging.

In the talk, an introduction to the Worldwide LHC Computing Grid from the
CMS point of view is presented, including a discussion of the
responsibilities of the Tier0, Tier1 and Tier2 centres for storage,
Monte-Carlo simulations and user analysis. Also the different grid
components used by CMS for data management on the grid and their interplay
are presented focussing on PhEDEx, a software framework developed for data
distribution on the grid.

To prepare the distribution and analysis of the first collision data of the
CMS experiment, several data and analyses challenges have been performed
during the last years. They have shown that dedicated monitoring systems are
indispensable for a stable operation. These systems are discussed with
respect to their usage at the German T1 centre in Karlsruhe, GridKA:
However, the large number of external monitoring systems combined with the
huge flood of information retard the identification of problems and
complicates the local administration. A meta monitoring system like the
HappyFace Project, which automatically queries the different relevant
monitoring systems, could provide a fast and comfortable access to all
important information, even for non-expert users.