System performance modelling WG meeting
Participants
- Local: Andrea Sciabà, Markus Schulz, Andrea Valassi
- Remote: Renaud Vernet, Michele Michelotto, Oxana Smirnova, Gareth Roy, Pepe Flix, Catherine Biscarat, Haykuhi Musheghyan, David Lange, Johannes Elmsheuser, Costin Grigoras, Concezio Bozzi
WLCG/HSF Workshop
Markus showed the tentative agenda of the two cost model sessions (see task list doc for details) and invited everybody interesting in giving a talk to volunteer. We will need also volunteers for taking notes.
In particular we need somebody to steer the discussion on Run3/HL-LHC, preferrably somebody who worked on the HSF community white paper. The talk on cost evaluation for sites could be given by Catherine or Renaud.
All comments on the agenda that prompted changes to the agenda itself are already reflected in the draft in the Google doc.
Task list update
Resource calculation
Andrea S. gave a summary of the meeting held yesterday on this task (minutes are attached to its agenda).
It would be very good, for the workshop, to show a simple set of code. David says that some of us are looking into how to approach this.
Example workflows
Concezio will make instructions available to run an LHCb workflow packaged as a Docker container that can be run anywhere and that runs all the processing steps from a shell script.
Cost estimation
Catherine started to reproduce cost estimates, she will complete it when she comes back from a leave of absence. Renaud pointed us to a document, asking for comments.
About the reference example given by Markus, he clarified that the amount of blocks read refers to the network and to the storage. David asked to add references to justify the numbers. Markus explains that they come from a 2000 events ATLAS RECO job with pile-up. Anyway what really matters is the calculation process, not so much the numbers in input.
Testbed
Pepe announced that he plans to put together a testbed at PIC for our activities.