Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 91: GRID computing

T 91.6: Vortrag

Donnerstag, 18. März 2021, 17:15–17:30, Tp

Implementation and benchmarking of a caching solution in the ATLAS Freiburg environment — Michael Böhler, Anton Gamel, Stefan Kroboth, •Dirk Sammel, and Markus Schumacher — Albert-Ludwigs-Universität Freiburg

The near future of particle physics will be an era of high luminosity: the High Luminosity Large Hadron Collider (HL-LHC), which is expected to be operational in 2027, will enable the collection of data corresponding to a total luminosity of around 3000 fb⁻¹ by the ATLAS experiment. In addition to this data, a large number of simulated events will be needed by a multitude of analyses. This results in the requirement for huge amounts of disk space. In order to cope with these enhanced storage requirements, scenarios with fewer, but larger storage sites and several compute sites without long-term storage are discussed.

Therefore, caching solutions are being developed. In this approach, the required data is transferred from external storage sites and then stored on local cache spaces. This enables fast access if the data is subsequently needed. The data is automatically deleted if it is not used for some predefined amount of time.

In this talk, the implementation of a caching setup, Disk-Caching-On-The-Fly (DCOTF), in the ATLAS Freiburg environment and the development of a benchmark that mimics a typical user-analysis are presented. The results of the benchmark testing the performance in Freiburg, especially in comparison to non-cached data access, are discussed.