MC productions as ALICE Service Tasks ALICE Offline

  • Slides: 9
Download presentation
MC productions as ALICE Service Tasks ALICE Offline Sep 10, 2012

MC productions as ALICE Service Tasks ALICE Offline Sep 10, 2012

Analysis trains • The common ‘alitrain’ account is the second-largest resources user presently 2

Analysis trains • The common ‘alitrain’ account is the second-largest resources user presently 2

Analysis trains (2) • Standard operation… • Pre-QM discussions • Split ‘alitrain’ in per-PWG

Analysis trains (2) • Standard operation… • Pre-QM discussions • Split ‘alitrain’ in per-PWG user • Migrate all to AODs • Track efficiency and resources usage separately, help improve both • Put on hold until after QM • Now we are ready to do it 3

Accounts • The PWG accounts are created pwg_pp pwg_cf pwg_dq pwg_ga pwg_hf pwg_je pwg_lf

Accounts • The PWG accounts are created pwg_pp pwg_cf pwg_dq pwg_ga pwg_hf pwg_je pwg_lf pwg_ud /alice/cern. ch/user/p/pwg_pp /alice/cern. ch/user/p/pwg_cf /alice/cern. ch/user/p/pwg_dq /alice/cern. ch/user/p/pwg_ga /alice/cern. ch/user/p/pwg_hf /alice/cern. ch/user/p/pwg_je /alice/cern. ch/user/p/pwg_lf /alice/cern. ch/user/p/pwg_ud • Job quota: 5000 unfinished jobs per PWG, CPU/Wall unlimited • File quota: 50000000 files, 300 TB • These are fully adjustable as needed 4

MC productions • Majority of the productions are PWG-specific • Exceptions are the ‘large’

MC productions • Majority of the productions are PWG-specific • Exceptions are the ‘large’ common production (see later) • Natural extensions of the PWG account split is to use these accounts also for the MC productions • Compatible with the “MC production request workflow” • Each PWG can run multiple productions, with internal prioritization of their choice • Many PWGs can run in parallel, within their quotas, the balance is assured by fair-share 5

MC productions (2) • The production macros and JDLs are by now standard and

MC productions (2) • The production macros and JDLs are by now standard and well tested • Production management framework (LPM) offers many options • anchor runs • trigger-weighted MC productions • RAW data embedding lists • The MC preparation and operation should be assigned as ‘Service task’ in the respective PWGs to students 6

MC productions - others • Common MC productions • Handled by PWG-PP, MC group

MC productions - others • Common MC productions • Handled by PWG-PP, MC group • The group will have its own account and ample quotas for large production requests • If necessary – higher priority and quotas wrt everything else 7

MC productions (3) • Extensive technical training is essential for the success of the

MC productions (3) • Extensive technical training is essential for the success of the schema • We can be ready with a course (including hands-on) in November 8

Summary • The migration to PWG-Service task MC productions will (in our opinion) have

Summary • The migration to PWG-Service task MC productions will (in our opinion) have several benefits • Faster production turnaround – after the test and QA, the production can be started immediately by the PWGs • Less chances for miscommunication and ‘stalled’ requests • PB approval workflow stays intact • Offline will organize the courses and support/extend the production tools 9