Attending:
- Alexei
- Simone
- Michel Jouvin
- Doug Benjamin
- Dirk Duellmann
- Brian Bockelman
- Giacinto Donvito
- Paul Rossman
- Gergely
- Maarten Litmaath
- Wahid
- Markus Schulz
- Philippe
- Elisa
Apologies
Minutes from last meeting:
Update from chairs
- See slides, no big questions
- From Ian Bird: can storage group and data management group come up with a joint "layered model" between SM and DM? Need volunteers.
- Note: There is an Data Federations Workshop happening next week. Brian, Dirk, Alexei, folks from ALICE (but not Maarten - Costin?)
- Will try to send a joint report from workshop to the email list
Input from ALICE
(Note: I just recorded the things I found interesting, or questions asked during the presentation. For the contents, see
WLCGTEGDataManagement_ALICE)
- Most data management is built on ALIEN file catalog. See presentations linked to aforementioned page.
- Think they can maintain the system for years to come with a few experts.
- FTS usage is really blocked by ALICE security envelope.
- View adding POSIX access to application as unnecessary complexity. Would be fine if Xrootd redirected to POSIX, as long as it is handled.
- PC: How is pledge information handled? ML: Have had a few cases where we filled up storage; most of the selection is based on network connectivity.
- ML: Find ourselves relaxing data placement constraints. Local access is preferred, but find it less important.
- PC: Are files copied around or streamed? ML: Streamed.
- Interesting discussion about maintaining CPU efficiency.
- BB: Average # of replicas? ML: Default is 3.
- Caching: we do this for conditions, not for physics.
- BB clarified: Initial intent is "physics data is pulled on demand and discarded automatically, without experiment involvement".
- DD: We need to elaborate several scenarios and ask about them explicitly.
Decided to delay the CMS presentation in order to not go overly long. Next meeting will be joint meeting with Storage in 2 weeks (on Thursday).