Request4: PWG3 - UrQMD, min. bias, BiBi @ 9 GeV

Monte-Carlo productions
1

We make a request for a production to be used for physics and methodological studies within our PWG3 in the scope of correlation femtoscopy.

The production is based on UrQMD Monte-Carlo input with min. bias events. The main idea of the production is motivated by our studies that we announced for this year within the Grant.

Briefly, it includes the following steps:

  1. To do precise tests of the current tracking procedure in TPC with available reco data in order to be sure it satisfies the minimal conditions on two-track effects (splitting and merging) and momentum resolution that look crucial for our studies.

  2. To perform a three-dimensional femtoscopy analysis that requires essentially more statistics to be used for.

  3. To finalize the structure and content of miniDst-format. It is used as input for our studies with the MpdFemto software that had been already tested successfully and now it has a pre-release state.

  4. To perform an analysis by making use of factorial moment technique. It also requires enough statistics in order to be able to estimate possible statistical and systematical uncertainties with regard to a final result produced by the technique.

We request to produce a dataset with 10M events . All necessary macroses
runMC.C (9.7 KB)
reco.C (8.0 KB)
are attached to the request.

For the simulations we suggest to use a “standard” default set of preferences. To avoid some mismatches in versions of macroses to be executed, as mentioned, they are attached to the request.

The production does not require to perform simulations with generator as input, since all necessary inputs are available in the mcDst-format. All interfaces to work with mcDst are ready and it allows one to avoid problems related to absence of space-time information on freeze out of particles (from generator), which is considered, indeed, as we must have.

Summarizing the request, we put here a preferable set of parameters for the production:

  1. 10 MEvents , UrQMD, min. bias, BiBi @ 9 GeV

  2. mcDst-input ( ncx.jinr.ru: /eos/nica/mpd/users/batyuk/mcDst/UrQMD/BiBi/Hg/cms_9GeV ). Each input contains 1 000 events.
    The naming convention looks as follows: BiBi_ecm9GeV_EoS_Hg_0-14fm_1000ev_ID.mcDst.root, ID pertains to a range [0 … 9999] without any skips done.

  3. A “fresh” version of the MpdRoot software in the dev branch (It has a revision number like fbe59ac) must be used . The used revision has to be tagged for future to avoid library mismatches in the case if being modified!

  4. Files produced after the runMC.C execution are considered as temporary ones to be used for reconstruction and we do not require to save them .

  5. When doing the production, one has to save both types of files to be produced as output: standard mpddst and miniDst.

  6. To validate the production configuration (output), we need about of 500 000 events reconstructed to be passed by our means through QA tests.

  7. The output files will have to be copied and stored at the NICA cluster if possible.

Reminder: Do not forget to compile the MpdRoot software in release mode ( set(CMAKE_BUILD_TYPE Release) in $VMCWORKDIR/CMakeLists.txt). It essentially speeds up the simu/ reco procedures.

In case of questions arisen concerning the production, please, contact P. Batyuk (pavel.batyuk@jinr.ru)

2

Is 1000 events per file required for the mpddst and miniDst files?

3

Hi,

The idea is to reduce the size of miniDst files as much as possible. In fact, more events in a miniDst file means better compression. But in this case, standard Dst files should not be too large. Therefore, the size of 1000 events is selected.
I think Pavel Batyuk can give You more details.

Thanks,
Konstantin

4

The problem is that the time limit for jobs on tier1,2 is 1 day. For 1 day, 1000 events on tier1,2 do not have time to be processed. I usually set 250 events per file. If we do not use tier1,2, then we lose about 950 cores. Production speed will decrease approximately 2.35 times.
If you agree to files for 250 events, I can do 4 tasks(4 dst and MiniDst files) for one input file.

5

Hi Andrey,

So, if the issue with number of events per one dst file is a question of time execution per one job, I think, you can do it in the way you have proposed.

6

Hi Andrey,

Thanks for explanations!
Yes, I agree with Pavel, it is OK to have 250 events per file.
Please, go ahead.

Thanks,
Konstantin

7

test sample 500 000 events:
mpddst:
lxpub:/eos/nica/mpd/dirac/mpd.nica.jinr/vo/mpd/data/exp/dst-BiBi-09GeV-mp07-20-pwg3-250ev/BiBi/09.0GeV-0-14fm/UrQMD/BiBi-09GeV-mp07-20-pwg3-250ev-1
lhep:/eos/nica/mpd/sim/data/exp/dst-BiBi-09GeV-mp07-20-pwg3-250ev/BiBi/09.0GeV-0-14fm/UrQMD/BiBi-09GeV-mp07-20-pwg3-250ev-1
MiniDst:
lxpub:/eos/nica/mpd/dirac/mpd.nica.jinr/vo/mpd/data/MiniDST/dst-BiBi-09GeV-mp07-20-pwg3-250ev/BiBi/09.0GeV-0-14fm/UrQMD/BiBi-09GeV-mp07-20-pwg3-250ev-1
lhep:/eos/nica/mpd/sim/data/MiniDst/dst-BiBi-09GeV-mp07-20-pwg3-250ev/BiBi/09.0GeV-0-14fm/UrQMD/BiBi-09GeV-mp07-20-pwg3-250ev-1

8

Hi Andrey,

Many thanks for such a quick response!
I need a couple of days to see whether everything is ok or should be changed / corrected.
So, as far as we finish the QA checking, you will be informed to go further.

9

Please check the files 4**-* more carefully. I processed them on the NIСA cluster. NIСA cluster was first used for the mass production using DIRAC.

10

Hi Andrey,

So, we did a quick glance on the produced data and it turned out to be some things to be corrected from our side. Namely, they are related to magnetic field and parameters of smearing of beam in the interaction point.

All necessary modifications have been done and they are in the dev branch of the repository (fa4d0bfe).
A macro to be used for simulations is attached to this message. The reco macro stays to be the same.

Please, produce not more than 200 000 events to be sure (from our side) that everything is correct. The previous data samples have to be deleted from the storages.

Thanks in advance!

runMC.C (9.5 KB)

11

OK. Will be done today.

12

Done. 200K events at the same place.

13

Ok,

Thanks a lot!

So, we need one day to check the data.

14

Hi Andrey,

We looked at the produced small data set you gave us.
It seems reasonable, so we do agree on that you can continue the execution of the production ordered.

Thanks a lot!

15

Andrey,

Please, do not start it immediately.
We decided to make an update for the dev branch.

16

Done.

Also, please, use a new version of runMC.C attached to the message.
runMC.C (9.5 KB)

17

And, also, remove, please, the events produced before.

–
Thanks!

18

OK. Today I will start 10,000,000 events. It is strange, tipicaly a notification comes with new messages in a topic. This time it was not. And another question, my eos storage on NICA cluster does not work. Do your eos folders work? I cannot copy anything to a folder on eos.

19

Hi Andrey,

I checked the point. I got something like that:

[batyuk@ncx105 batyuk]$ eos mkdir test
error: unable to create directory (errc=28) (No space left on device)

despite my quota is not reached yet .

20

Thanks! Eos is broken again …