02 Plan of action
W 2012 task force January 3 , 2012, updated March 12
Parallel directions of analysis:
* 2009 muDst : reply w/ new W algo
* 2011 muDst : run QA, EMC calib check (ped, stat, gains -> Alice O.), spinBits upload to DB, replay W algo, vertex QA (pileup), estimate not reducible QCD bckg, vet stability of AL & false asymmetries
* 2012 : oversee data taking (+cdev, +L2 mon) , real time W reco at a remote site, stage daq-files for BFC production, help with TPC+vertex calib - if needed, follow all steps for 2011 muDst
* write W AL paper now : assume we will get results similar to the those on the cumulative AL projection plot. LHC can publish weeks after their run stops, we can do it months after run stop too.
* simulation/embedding needs?
- Yes, for W->tau and Z->ee backgrounds (maybe use Run 9 samples?), need estimate of absolute lumi for norm. but not vernier scan.
- No, for QCD unless study of veto by clusters in the 1st FGT disk.
Technical challenges:
- reserve, customize, and maintain CPU power for real-time BFC production over pp200+pp500 2012 data taking.
- stage simultaneously W-muDst from 3 years at one computing facility, allow for 2 different copies for run 2012
- assure sufficient CPU to replay all W-muDst within one week.
- stage all W-daq files from 2012 to await BFC production with post-run STAR calibration
- run post-run BFC production as 1-2K parallel jobs, assure muDst get registered at RCF
My guess is we will need 200 CPU x4 months for the real-time processing, later 2000 CPU for 1 month for BFC production of full 2012 W-data set. Disk space: 100-500 TB - if we manage to avoid partial HPSS storage.
Tasks below require a dedicated person per task, parallel effort. Not all tasks will last the same amount of time nor require the same daily effort. Writing of the needed/missing software not included.
- data transfer, job submission, monitoring of CPU load - 4 months of 4 hours/day job, could be more
- prepare, maintain STAR code at a remote farm: at start 1 month full time, later 1h/day for 4 months
- QA of reco Ws - 4 months of 1h/day job, may be combined with QA of L2 plots
- QA of 2011+2012 STAR runs - many months of 'desk work'.
- determination of spin bits loaded to DB for run 2011+2012 - 3 weeks of work after muDst are available.
- BEMC calibration for 2012 - a month of work with muDst (use Zs?, should find ~100), assuming HV not changed, plus watch L2ped during pp500 run
- EEMC calibration for 2012 - gain likely lower than Run 9 and HV was changed to correct , so new calibration needed (MIP, pi0, Z?)
- writing STAR analysis note of the progressing analysis (many people describe their work, one supervisor is needed)
- TPC calibration+beam line position - I hope STAR Calib team will do it all?
- TOF calibration to improve PPV vertex quality - status not known
Organization:
- use only Drupal to document progresses.
DONE: head-page w/ edit4all privilege http://drupal.star.bnl.gov/STAR/node/10125
- weekly dedicated phone meetings (which you should not miss)
- one 3-days face-to-face workshop before Summer vacations
- Jan will be custodian of master latex draft of the W2012 AL paper
- Justin will be custodian for latex draft of W2012 analysis note
- Gene will oversee progress of calibrations
- Hal will oversee QA of runs, Justin will generate raw .cvs table, Jan will do first pass of automatic run rejection
Boundary conditions,
understood as follows: Everyone is free to do more than the minimum we agree is our goal, but should not delay delivering of his/her share of the work because of such additional studies.
* focus on releasing cumulative preliminary results for W AL as part of the paper draft presented to GPC in early Fall of 2012. Avoid earlier partial releases to not burn our resources.
* Endcap Ws only form data taken in 2011+2012
* only TPC tracking used to reco Ws - this limits the eta range to [-1.3,+1.3]
* no use of FGT unless significant benefit of veto of QCD is demonstrated w/ M-C for W eta>+1. Only 1st FGT disk, cluster based, no use of FGT tracking. FGT 1st disk must perform well in run 2012.
* no A_N for Ws from pp500 transPol run form 2011
* no x-section determination for W - not care about Vernier scan for 2011, 2012 data.
* no study of W+/W- x-section
Brake down of selected tasks
====== run QA for year 2011, 2012=====
Decisions/comments:
- are we doing work for the whole spin PWG or narrow it down to W AL paper?
- some of the steps may differ year-to-year, we should strive for unification of cases
- can we use 2009 run list as is, or some runs needs to be rejected based on Ross QA of vernier scan?
Steps:
- review list of relevant problems (key-words) of STAR detector elements
- review list of variables per run extracted automatically from DB (the online page)
- define (minimal) check sequence for every run, where are the plots?
- automatically generate full 'possible-good' run list, reject automatically as many questionable runs as possible.
- divide runs in to several segments and assign QA people
- name QA progress supervisor, who will track progress (also for ongoing 2012 run) and summon those who stale
- merge inputs from QA-people in to the master spread sheet.