System performance modelling WG meeting
- Local: Andrea Sciabà
- Remote: David Lange, Renaud Vernet, Costin Grigoras, Johannes Elmsheuser, Michele Michelotto, Gareth Roy
Renaud is preparing a questionnaire for Tier-1 sites to inquire about their cost parameters. He got comments from Catherine and Alessandro. In a meeting in July it was proposed to ask the sites the cost of replacing a certain amount of disk with tape, but Renaud thinks that this information would not be particularly helpful because it does not say anything about the evolution of costs. Renaud will ask for more feedback on the questionnaire (for example to Pepe and Oxana).
About the CHEP paper, it was decided to sue Github for the collaborative writing. About Overleaf, it seems that the CERN licence is available only for people working from a CERN IP address, so it is not an option.
About whether having a pre-GDB or a session at the HSF/WLCG workshop in March, the consensus is that we should ask for a pre-GDB early next year, which would be also a good milestone.
About the collaboration with DOMA, we should try to feed information in both directions.
Action: review the glossary, in particular document the workload names (it is very useful for students).
About the reference workflows, people who have run them are invited to share their experience to make sure that issues are well known.
Action: check that the workflows are still current.
Markus mentions that Corentin (a student collaborating with us) has worked to measure the sensitivity of workloads with respect to latency and bandwidth resctictions. His code will not be included in Prmon but will be made available elsewhere. The work on time series parametrisation, however, did not produce significant results.
David L. will share his new version of the resource estimation framework of CMS.