Difference between revisions of "GDB 13th January 2010"
Latest revision as of 20:40, 13 January 2010
Grid Deployment Board Notes 2010-01-13
N.B.: As with many others I missed the start of the meeting because of 20cm of snow overnight. I had to join by EVO in the morning.
Talk summarised improvements in GGUS system (improved availability).
More frequent tests of alarm system, now managed by GGUS team. Some discussion of how often (once a week, regular is too often, but should test after changes to system).
Main issue has been about multi-user pilot jobs. Maatren summarised the issues and discussion so far, this is the useful URL:
Philippe Charpentier proposes links to descriptions of the experiment frameworks so that people can understand what is involved. Much more worried about data security. Considers
Are we fudging established security policy (Steve Newhouse)?
PC - policy was decided without scrutiny of frameworks. Why no review of middleware itself?
Kors - in favour of fixing all security holes; but most worried about putting extra hurdles in the way of analysis at a critical time. Maarten/John - this supports option 4, while gaining experience.
Michele Jouvin proposes reviewing questionnaires in a pre-GDB.
In addition: other working groups are proposed, related to data management in the main.
Some feedback on user registration - often good suggestions which have been incorporated and helped to simplify and generalise the policies.
New policies for AUP and Site Registration have been circulated. Comments by end of January.
Proposal to drop gLite 3.1 two months after gLite 3.2 version has been demonstrated to be 'good working service'.
I asked about the impact on code portability by encouraging a development/build monocoluture. Oliver and others thought there was no so great an impact. There seems to be little demand for gLite on other platforms.
CMS would like a 32 bit version of the UI on SL5.
Proposal to rationalise the meta package versions to be more meaningful.
Updated on OSG activities, mainly within WLCG context.
Prioritised SL5 (almost done) over CREAM CE (not done).
Basically happy, but concerns about CASTOR at CERN. Need improvements in SRM monitoring and would like to see discussion about future of storage.
First talk from new CMS computing coordinator, Ian Fisk. Generally happy with computing operations. Some discussion about requirements on WNs for cache space - 20GB per core. On high N core machines will hit i/o limits.
Detailed list of problems seen, but nothing catastrophic. Still forced to download data to WN at dCache sites because of some dcap bugs.