Distributed Computing Activities December 2011 ATLAS Distributed Computing
Distributed Computing Activities December 2011 ATLAS Distributed Computing Issues ADC Weekly, 20 December 2011 Alexei Klimentov BNL December 20, 2011. A. Klimentov
Completed Grid Jobs (20. 11 -20. 12 2011) Distributed Computing Activities December 2011 Jobs Completed jobs. Day average MC simulation Grid Analysis CPU 10^9 sec CPU consumption. Day average December 20, 2011. A. Klimentov 2
Distributed Computing Activities December 2011 Tier-0 Operations. December 2011 l Tier 0 CPU utilization is ~100% 5000 job/slots n Initial Plan (Dec 14 th) : Ø Ø Physics Hard. Probes : l 3. 5 M events 1. 2 Days Physics_Min. Bias l 38. 6 M events 6. 4 days l Older run reprocessing and reprocessing with different reco cuts will be organized in parallel Start reconfiguration HI processing December 20, 2011. A. Klimentov CERN LSF issue during weekend 3
Distributed Computing Activities December 2011 Physics Coordination Request l Kevin Einsweiler & Richard Hawkings Ø As we are progressing in completing the MC 11 samples for 2011 analyses, and beginning to launch first MC 12 samples to prepare for trigger and reconstruction in 2012, it is very clear that the present excellent performance of Tier 1 and Tier 2 will be critical for our success in the coming months. However, we have built up an enormous backlog of MC 11 jobs (both G 4 fullsim and digi+reco) that will require all of our CPU resources to clear out before winter conferences and before we have to switch to MC 12 to prepare for 2012 collisions. This is in part driven by the change in machine energy to 8 Te. V foreseen for 2012, and in part by the larger MC samples required to analyze 100 times more luminosity in 2011 compared to 2010. We therefore wanted to alert you to the critical need for high availability of our grid computing resources throughout the yearend period, to the extent that this is possible. . . December 20, 2011. A. Klimentov 4
Distributed Computing Activities December 2011 Distributed Computing Operations. Dec 2011 l Requests received by Distributed Computing n Monte-Carlo Ø Priority 1 samples (digi+reco) Ø G 4 priority 2 samples Ø Possibly also priority 0 (if the pile-up setup will be changed to Pythia 6) l Ø TBC by Physics Coordination Tasks are submitted in ‘blocks’ to avoid ‘summer effect’ l ~13 k MC tasks were submitted during last week n Heavy Ion data processing n HLT processing at CERN Ø Until 12/23 and after 1/03 Ø More activity in January is foreseen n Support Grid analysis, group production and data distribution to sites December 20, 2011. A. Klimentov 5
Distributed Computing Activities December 2011 Distributed Computing Operations. l The first priority is to run MC production l ATLAS Computing Facilities at CERN Ø Give as much CERN resources as possible to the Monte-Carlo Production l J. Boyd, A. Di. Girolamo, A. Nairs, AK n Rearrange ATLAS Tier-0 resources n 2500 cores will be given to Grid processing n ~700 cores will be kept to continue HI processing at CERN n Assignment is done until Jan 4 th, the situation will be reassessed in January. Ø HI processing at CERN will be continued with ~700 cores December 20, 2011. A. Klimentov 6
Distributed Computing Activities December 2011 ATLAS Computing Facilities at CERN MC Production@CERN l Grid Analysis@CERN ATLAS share was reconfigured today (Tue Dec 20 th , Alessandro Di. Girolamo, Gavin Mc. Cance) Ø Current configuration : l Grid prod: 3 k on public, 1. 2 k on atlast 0 l Grid analy: 1. 5 k public 0. 9 on atlast 0 l Tier 0 : 0. 1 on public, 0. 5 on atlast 0 December 20, 2011. A. Klimentov 7
Distributed Computing Activities December 2011 Distributed Computing Operations. Support. l ATLAS Computing Facilities at CERN n Best effort support from CERN n Distributed computing expert on-call (AMOD) support (except Xmas and New Year) Ø The first line of support l n Best effort support from ADC experts Ø The second line of support n Central services at CERN Ø Can be contacted via CSC phone in case of problems ATLAS Grid centers n n ADCo. S shifters will provide the first line of support Ø Usual coverage, except Xmas and New Year Ø Critical issues will be escalated to AMOD (and if needed to CERN-IT and ADC experts) Best effort site support on some of them and full support on the others Ø ADCo. S and AMOD will follow Grid sites availability December 20, 2011. A. Klimentov 8
Distributed Computing Activities December 2011 Happy New Year ! l Many thanks to all Distributed Computing people, shifters and sites for the fantastic work during 2011. Fabiola Gianotti : “Limitation to release results is not with computing “ December 20, 2011. A. Klimentov 9
- Slides: 9