The ATLAS production system management monitoring Th

The ATLAS production system management + monitoring Th. Naumann

Computing Operations • ATLAS CTDR: – 200 Hz and 320 MB/sec – 17 Mevts/day @ full effi. – MC: 20% of raw data rate – 3 Mevts/day MC • T 1 s: calibration, alignment, reconstruction, reprocessing • T 2 s: MC production world-wide with >100 sites – – 2 min/evt: 50 evts/job (physics events) ~100 kjobs/day ~ 1 Hz submission time now: ~ s analysis up to ~1 Mjobs/day: job rate 12 Hz (CMS: n*100 K jobs/day) • needs MONITORING !

Prod Sys Don Quijote Eowin OSG Lexor I, ES Condor. G ES, F, D CERN Dulcinea Panda OSG • • • prod. DB: unique, records job execution, definition, logical files supervisor: – – – give job from prod. DB to executor, but no brokering ! follow job status till done check output + update file tables executor: – – interface to specific Grid or batch systems stage files in/out

Computing Operations central teams at CERN: • CERN T 0 operations (Luc Goossens/CERN) • Job executors (R. Walker/TRIUMF) • Service Challenges SC 4 (Gilbert Poulard/CERN) • Worldwide distributed production (Kaushik De/UTA) • Site+software installation+validation (A. De Salvo/Rome I + J. Kennedy) • Job definitions (Pavel Nevski/BNL) • Database replica deployment (Alexandre Vaniachine/ANL) • DDM operation + software integration (Alexei Klimentov/BNL) • Production monitoring (John Kennedy/Univ. Munich) • User support (Guido Negri/INFN-CNAF)

Zeuthen Prod. Sys site info:

Production System Monitor John Kennedy, LMU Munich: http: //www. etp. physik. uni-muenchen. de/~jkennedy Monitor: http: //atlas-php. web. cern. ch/atlas-php/Db. Admin/Ora/php-4. 3. 4/proddb/monitor/Job. Info. php • executor efficiency • 20. -21. 4. 24 h finished 5833 , failed 3197 : efficiency 65%

CPU/day ￭ LCG-CG ￭ LCG ￭ Nordu. Grid ￭ OSG k. SI 2 kdays Prod. Sys Jobs/day ￭ finished ￭ failed