Speaker
Dr
Doris Ressmann
(Karlsruher Institut of Technology)
Description
All four LHC experiments are served by GridKa, the German WLCG Tier-1 at the Steinbuch Centre for Computing of the Karlsruhe Institute of Technology (KIT). Each of the experiments requires a significantly different setup of the dCache data management system. Therefore the use of a single dCache instance for all experiments can have negative effects at different levels, e.g. SRM, space manager and metadata (PNFS).
For reasons of performance, operation and administration, GridKa has started in 2008 to prepare to split the dCache storage management system into four instances, each serving a single VO. First, after stringent planning, the ATLAS VO is migrated to a dCache instance. Several checks are made to ensure consistency between physical and logical data contents, because a total of three different databases are involved in the conversion. The procedures, the tools which have been used and the planning for the ATLAS dCache fork are described, followed by first evaluations of the split environment. Also, the realisation of the different computing models into the current dCache setup at GridKa is presented.
| Presentation type (oral | poster) | oral |
|---|
Author
Dr
Doris Ressmann
(Karlsruher Institut of Technology)
Co-authors
Mr
Artem Trunov
(Karlsruher Institut of Technology)
Dr
Christopher Jung
(Karlsruher Institut of Technology)
Mr
Jos van Wezel
(Karlsruher Institut of Technology)
Dr
Silke Halstenberg
(Karlsruher Institut of Technology)
