Download presentation
Presentation is loading. Please wait.
Published byMerryl Whitehead Modified over 8 years ago
1
ATLAS intentions on Job Priorities Simone Campana
2
Job Priorities TCG WG Job Priorities a)Computing Element and Batch System: VOMS FQANs correctly mapped to proper UNIX groups Shares defined at sites (T2: 50% prod, 50% analysis) for given UNIX groups b)INFORMATION SYSTEM (BDII): Infos published with breakdown per FQAN (VOViews) c)WMS: Needs to digest VOViews properly in order to schedule jobs Successfully tested in INFN (PBS sites) Need for (a), (b) and (c) depends on Production and Analysis frameworks
3
ATLAS would like to start using Pilots for analysis everywhere – In parallel to WMS – See Graeme’s talk Using pilots for production and WMS for analysis – Need surely (a) – (b) would be needed to help the WMS making a decision about “emptiest” queues for ANALYSIS Using pilots for production and analysis – Need still (a) at least if you send “Production pilots” and “analysis pilots” – No need of (b)
4
ATLAS would like to ask deployment of TCG Job Priorities everywhere, following the model prototyped in France – This introduces a the Pilot Role as in LHCb For (a), define three shares for ATLAS *PLEDGED* resources Production (/atlas/Role=production) (80% in T1, 50% in T2) Analysis Pilot Analysis (/atlas/Role=pilot) Direct Job Submission (/atlas) Up to clouds to decide analysis shares Starting point for EGEE sites: 50% Pilots, 50% direct job submission OSG support pilots only NDGF in the process of migrating to production via pilots. A decision for analysis will be taken later – Possibility to deploy further shares for local (*UNPLEDGED*) resources /atlas/fr Sites should deploy (b), unless they do not offer share for Direct Analysis
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.