DOMA / ACCESS Meeting
People at CERN: Markus Schulz, Andrea Sciaba, Riccardo Di Maria
People remotely: Diego Ciangottini, Horst Severini, Johannes Elmsheuser, Martin Gasthuber, nhardi, Nikolai Marcel Hartmann, Stephane Jezequel, Marc Weinberg
Stephane Jezequel : Next meeting after CHEP - November 19th, 2019
Markus: this presentation follows the work performed by former summer student and already presented in DOMA ACCESS
For details, see slides. Clarifications:
- regarding “cost optimisation”, it is not a general estimate, but ATM really depends on site and data used
- SW releases could affect the “probability based on the number of accesses and time since last access”? unlikely but would be interesting to check
- does the structure change per data type? the shape parameter could differ in principle
- same pattern observed at CERN - to be added to the slides
- failed jobs could affect the shape parameter but this would mean a high job fail rate
The plans are to combine these studies with ATLAS data Rucio record. Problems from different modus operandi implemented by the two experiments. But other metrics could be used for studies.
Studies on ATLAS already ongoing on different sites.
- ATLAS Ntuples from DAODs (10-50/100 kB/event) are roughly 1-10 kB/event but really analysis dependent