Difference between revisions of "GDB June 2008"
(No difference)
|
Latest revision as of 09:03, 14 July 2008
== 11 June 2008 == Agenda
Contents
UK input/issues
Meeting summary/report
Note: CERN was not broadcasting video for the morning and first afternoon talks, which made it hard to follow the meeting as you could not see what slide the speaker was on.
I have only really noted discussion points - see the agenda for slides.
Introduction
July will not be a Tier-2 GDB. Probably September instead. This would allow T2s to prepare any questions / presentations beforehand - I suspect this should be raised at next tb-support meeting?
Security
It was noted that there was a proceedure for individual grids to report exceptions to general policy.
Pilot jobs can be submited by "robot" certificates, but this requires VOs to name a real person who is responsible.
Comment deadline is next Wednesday. The documents will go to WLCG MB for final approval.
I missed the chance to ask about VO responsibilities when users break the AUP - I will email DK.
CCRC08
Jamie summarised points for discussion in next day's post mortem workshop. "We are ready to face data taking..." Still have to find a way around emerging problems. Will focus on discussion.
HEPiX
Quick review of last meeting at CERN. You can see all the slides in indico:
http://indico.cern.ch/conferenceDisplay.py?confId=27391
Benchmarking: There will be a standard benchmark which we can use, but not quite obvious which one will scale best and be the most useful to use (some people thought they were all much of a muchness). Previous problems with scaling of HEP code vs. benchmark seem now to be understood. Dario noted that the ATLAS G4 simulation gets different results when run on AMD an Intel (statistically the same though).
Pilot Jobs
N.B. Focus is on multi-user pilots. Security for single user pilots is better understood.
IB asked ML to come up with concrete recommendations to the MB about which frameworks are acceptable.
Noted that glexec needs better documentation.
Storage
Note that developer effort will be lost to bug fixes when real data arrives and this will affect implementations.
Will need to revisit long term goals in the light of first data taking.
Storage Accounting
Much work needed to be done to ensure data published and presented is good.
Job Efficiencies
CERN system for measuring efficiencies of all jobs independently of LSF accounting. In principle can be adapted for torque.
OSG
Some discussion on different policy on logging (no OSG req. for 90 days of logs). Site names are handled manually for now, which leads to odd naming problems.
