München 2009 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 68: GRID Computing 1
T 68.5: Gruppenbericht
Montag, 9. März 2009, 18:05–18:25, M110
ATLAS Distributed Data Management and Distributed Analysis test in the German cloud — •Cedric Serfon1, Günter Duckeck1, Johannes Elmsheuser1, John Kennedy2, Simon Kirichu Nderitu3, and Rodney Walker1 — 1Department für Physik, LMU München — 2Rechenzentrum Garching der Max-Planck Gesellschaft — 3Physikalisches Institut Uni-Bonn
To deal with the huge amount of data which will be generated by the experiments at the LHC, computing grids have been set up. These grids allow us to split the computing resources as well as the data over various computing centers. Grids are subdivided into smaller structures named clouds that group around a big regional computing site, called Tier-1, smaller sites called Tier-2s. An important aspect of Grid Computing is the distribution of data amongst the sites and clouds. In ATLAS this Distributed Data Management (DDM) is a complex system that is built on several services (FTS, SRM...) and catalogues (LFC, DQ2 catalogues). Many tests of DDM have been conducted to check the performance of the system and will be described. Procedures and tools developed to ensure a stable operation of DDM in the German cloud are also detailed. In parallel to this activity, stress tests for Distributed Analysis are performed : Multiple bunches of jobs have been submitted on every cloud to check that users will be able to access and run efficiently on the data distributed by DDM. Many observations have been made during these tests, leading to a better tuning of the distributed analysis tools and the storage systems.