Home » Groups » STAR Protected

Test_Production_runList_Selection_Criteria

Updated on Mon, 2013-12-23 14:24. Originally created by devika on 2013-11-18 09:51.

 

# of Qa d' runs used from the First priority list 1709
run period  03/23/2013 - 06/10/2013
   

**************************************************************************************************************************

Golden list selection steps:

step category # of runs # of excluded runs 
1) No scaler runs (Qs) 107 107
2) TPC laser runs (Ql) 291 42
3)  Anomalous bunch crossing (Qx) 140 140
4) Run stopped due to     
       a) BEMC problems (Db)  7  most (Double checked)
       b) EEMC problems (De)  11  some (Double checked) 
       c) various STAR trigger problems (Dg)  41  some (Double checked)  
       d) various STAR detector problems (Do)  81  some (Double checked)  
       e) STAR daq rates problems (Dq)   15  most (Double checked)
       f) polarization measurement  44   None , except with other reasons(Double checked)
       e) TPC problems (Dt)  38  38
 5)  Major TPC, BEMC problems (Cx, Bx)   1,1  1,1
 6)   Major EEMC problems (Ex)   78  78 (Double checked)
 7)  STAR trigger problems (Tr)  148 most (Double checked)   
 8)   runs with no BAJP, JP0*Dijet, JP1*Dijet  12  0

*****************************************************************************************************************************************

General remarks of runs which are included

1)  Any run with small missing regions in TPC charge per pad plots . 
Ex,

2) Any run with small BEMC missing region
Ex,
 
3) Any run with EEMC hot region
Ex,
 

4) Any run with STAR detector problems/errors (TOF,FGT,FMS,MTD,VPD,ZDC,BBC) except TPC,BEMC and EEMC unless if the issue is so huge which caused run to be unacceptable 

5) Any run with STAR trigger  problems except TPX,BTOW,DAQ  unless if the issue is so huge which caused run to be unacceptable 

**********************************************************************************************************************
Note:

 
 
Golden list Day82-Day101    # of runs= 150
Golden list Day102-Day128   # of runs= 327

criteria:

Here is the BFC chain for reconstruction that we will want to use (no FGT):
"DbV20131223 pp2013a mtd btof VFPPVnoCTB beamline BEmcChkStat Corr4 OSpaceZ2 OGridLeak3D -hitfilt"

Looking at the "golden lists", we have 150 runs from days 082-101, and 326 from days 102-128, for a total of 476 runs. So the plan will be to do one file from each run, 1000 events from each file, which is 476,000 events. The resources necessary then are:

  • Processing time: it takes ~1 CPU-minute per 1 event, so this is ~17 hours per job (1000 events). The total will be ~476k CPU-minutes = 330 CPU-days. If we are able to run ~100 jobs per day, this will take us ~1 work week.
  • The output MuDsts use ~750 kB/event, or ~750 MB/1000 events, so ~360 GB total.
  • The input DAQ files use ~1.4 MB/event, but we will need to place on disk files that generally include more than 1000 events. I see on average ~2400 events/file, so 476*2400*1.4 MB = 1.6 TB for DAQ files. If we stage ~100 files per day, and erase them after their jobs run, then we can trim this down to ~400 GB of space if necessary (i.e. expected space does not become available). Note that we have 82 files on disk from days 139-146, but these are not of interest currently.

We *must* practice read once / write once on these disk to avoid overloading the NFS servers. In essence, we copy input DAQ files to scratch space at the beginning of the jobs, run jobs there, then transfer output MuDsts to central storage space only at the end of the job.

Jobs can be run by both Devika and me. We need write access to the storage space. At the moment, some of the directories are there, but we do not have write access to any of them.



Groups: