Aachen 2019 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 59: Grid-Computing und Software
T 59.2: Vortrag
Mittwoch, 27. März 2019, 16:15–16:30, S11
Tuning ROOT parameters to optimize ATLAS Analysis Object Data — •Martin Errenst1,2 and Attila Krasznahorkay1 — 1Bergische Universität Wuppertal — 2Cern
In ATLAS, reconstructed physics objects are stored in the centrally managed xAOD format. xAOD is based on ROOTs file format and therefore offers a range of configuration settings. In a study where typical file access patterns in ATLAS analyses and other workflows are emulated, these settings are reviewed to optimize for small file sizes and good reading speeds.
The results of this study will be used to find optimal storage configurations for datasets used in different stages of ATLAS analyses. Finding good middle grounds between file size and file reading speed for different requirements of the final stage analyses and that of central data processing, is becoming even more important as ATLAS is preparing for LHC's Run3.
As a cross check to the analysis emulation, similar computing performance studies are done in an ongoing SUSY analysis, which is searching for R-parity violating decays in final states with one lepton and many (>8) jets of which 0 or >3 are b-tagged.