Notes from the first RAC meeting 25 Jan., 2006.
Below are the meetings notes from Rob merged with my notes:
Jim, Howard, Mike, Bruce, Bruce Bing, Srini,Kaushik, Rob
Jim:
Worry about policy implementation later.
Tier1 and Tier2 resources must be allocoated by this committee.
There are other resources around as well. This committee should issue guidelines for those.
Activities: managed production and individual users.
Kaushik:
State and plans for production.
3000 CPUs needed right now - ATLAS-wide.
25% for US.
11.0.4 samples B, C
5M high priority events before rel 12 comes out.
How long do we need these? 5-6 weeks.
BNL: 550 CPUS, 200 allocated to production
UC: 128
UTA: 150 --> 500
BU: 120
We have about 1000 CPU direct US ATLAS resources. We need to get the storage number for next meeting.
Jim: plots of usage for meetings would be useful.
Who are the big users?
Sanjay, who else?
What are needs for private production in the next few months?
Policies for database accesses - private vs managed production.
My Tier2 concerns:
Formulating a clear policy for usatlas resources for
- production team via panda
- private analysis via panda/dial
- private analysis via grid
- private analysis via local job submission
- private analysis via interactive use
Storage
- custodial storage via DDM
- physics group-level datasets
- private datasets
Accounting and auditing
Facilities need to publish this information in a consistent way.
Source LCG production
Non USATLAS production on OSG
Storage and CPU capacities at every site.
Planning details for datasets.
How much storage do we need via SRM/dCache.
Next meeting - focus on implementation
- role based authorization
- implementing priorities
Action Items:
- Jim will draft a mail to send out to the collaboration asking for planned usage (CPU and storage)
- We need a consistent way to publish policy:
Srini will setup wiki page on RAC.
(done:
ResourceAllocationCommittee)
Next meeting in ATLAS week: 23 Feb. 2006 Noon EST. We will use the UM video room at CERN.
--
SriniRajagopalan - 30 Jan 2006