Home » PANDA » PANDA - Computing » Grid and Infrastructure » Upcoming DC
|
Test Jobs [message #6325 is a reply to message #6203] |
Wed, 09 April 2008 17:06 |
|
Dear all,
In the test job JDLs please make sure they save output (if the case) to a directory writable by the production user 'pbarprod',
even if you try the job as yourself or as 'pbartest'. We will use 'pbarprod' for the data challenge next week.
Of course, I will doublecheck the JDLs proposed for the DC.
|
|
|
|
|
UrQMD_Smm [message #6333 is a reply to message #6325] |
Wed, 09 April 2008 22:01 |
Johan Messchendorp
Messages: 693 Registered: April 2007 Location: University of Groningen
|
first-grade participant |
From: *xs4all.nl
|
|
Hi,
I installed a 32 bit and a 64 bit version (binaries with libs only) of the UrQMDSmm eventgenerator on the Grid, compiled against root5.18 (which version is the most actual one for the external packages). It seems to work for the sites I could test it on. To test it on any site:
Installation (presently already instaled at KVI,GSI,Juelich,Dubna,Bucharest)
---------------------------------------------------------
packman install pbarprod@urqmd::r2008
test scripts
---------
/panda/user/p/pbarprod/bin/testurqmd.sh
/panda/user/p/pbarprod/jdl/testurqmd.jdl
i.e.
submit /panda/user/p/pbarprod/jdl/testurqmd.jdl <id> <momentum> <nrofevents>
will do the job with the output @
/panda/user/p/pbarprod/jgm/urqmd/run<id>/..
Johan.
(ps thanks to Aida and Vladimir!)
[Updated on: Wed, 09 April 2008 22:38] Report message to a moderator
|
|
|
Storage@KVI [message #6334 is a reply to message #6203] |
Wed, 09 April 2008 22:13 |
Johan Messchendorp
Messages: 693 Registered: April 2007 Location: University of Groningen
|
first-grade participant |
From: *xs4all.nl
|
|
Hi,
I have made 2TByte storage space (temporarily) available at KVI in case needed for the DC. This space should already be accessible via the SE.
Johan.
[Updated on: Wed, 09 April 2008 22:19] Report message to a moderator
|
|
|
|
one more week left [message #6346 is a reply to message #6325] |
Thu, 10 April 2008 16:11 |
Paul Buehler
Messages: 11 Registered: October 2007
|
occasional visitor |
From: *smi.oeaw.ac.at
|
|
Dear colleagues,
I prepared a schedule for the days remaining until the start of the Data Challenge (summarizing what has been discussed so far). It is displayed on http://panda-wiki.gsi.de/cgi-bin/view/Computing/DataChallenge1#Schedule. Inputs, comments are highly appreciated. I will try to coordinate the final steps to be able to run the jobs on 17.-19. April.
The packages will have to be finalized very soon, installed and testd on the sites, and we have to decide on the macros to run. As a first step I would like to grasp the current status. Could you please help with this?
* Packages:
Can you Soeren, please summarize the status of the software. What is the current status, what is left to do?
* Macros:
The latest proposal by Soeren was (or did I miss something?)
--------------------------------------------
stt
===
root -b runsim.C"(nEvents,pT)"
root -b rundigi.C
root -b runreco.C
pT=30,40,50,...100 MeV/c
pT=100,200,300,...,1000 MeV/c
pT=1,2,3,...,7.5 GeV/c
10,000,000 events each
tpc
===
root -b run_sim_tpcmvd.C"(nEvents,pT)"
root -b run_rectrack_tpcmvd.C
pT=30,40,50,...100 MeV/c
pT=100,200,300,...,1000 MeV/c
pT=1,2,3,...,7.5 GeV/c
10,000,000 events each
dpm
===
Events and beam momenta for dpm will be proposed tomorrow
--------------------------------------------
Soeren, do you want to update?
What about the UrQMD and the fast sim macros Johan has already installed and tested on the grid?
Could you, Soeren and Johan, please try to make a concise list of possible simulations to run, based on the packages which will be available for the DC01?
Can you Johan please summarize the status of the macros and the jdls. Examples for UrQMD and the fast sim are obviously ready. What about the other simulations (stt, tps, dpm)? Will you be able to take care of this?
* Storage:
One event is 250 Bytes (the number was given by Soeren). Is this true for any of the prposed simulations (UrQMD?)? If yes then we most probably have far enough storage capacity (1TB ~ 4.E9 events).
* CPU time
Can somone give a (rough) estimate of the CPU time used per event? This number would help to decide on the number of events to submit (per job, total).
Cheers,
Paul
|
|
|
|
Re: one more week left [message #6356 is a reply to message #6346] |
Thu, 10 April 2008 18:32 |
Jens Sören Lange
Messages: 193 Registered: June 2005
|
first-grade participant |
From: *physik.uni-giessen.de
|
|
Hi Paul and all,
Quote: | * Packages:
> Can you Soeren, please summarize the status of the software.
> What is the current status, what is left to do?
|
I can only comment on PandaRoot itself.
There were only 2 issues left.
1.) if riemannfit is on, then lhetrack cannot run.
This is more complicated and will need time to solve.
However, lhetrack is global (has tpc and mvd),
riemannfit is local (tpc only) in our understanding
So we decided to switch riemannfit off for the DC
(I am sorry to Sebastian and Tobias for this,
but we just have to make a decision just due to time pressure).
2.) include mvd into the tpc macros
(and so run the _global_ tracking for both mvd and tpc).
This was solved today by Stefano with help of Tobias and Ralf.
He checked it in.
-> this means:
rev---- is final for tpc, stt, dpm.
As the UrQMD code is not in the svn repository, this means:
rev2480 can be regarded as the PandaRoot version for the DC
(note:revised by Soeren on Friday Apr 11 15:37)
(only exception: if there is a last minute bug fix)
THIS MEANS:
(please read and enjoy)
we will run global(!) tracking for tpc digis(!) and mvd digis(!)
*Quote: | Macros:
The latest proposal by Soeren was (or did I miss something?)
[...]
Soeren, do you want to update?
|
Yes, I would like to propose the beam momenta for the dpm simulations.
(these are input parameters for the bash scripts prepared by Johan).
[p_beam / GeV/c][what for?]
0.739 (for PhiPhi at 2.000 GeV)
2.202 (for PhiPhi at 2.500 GeV)
4.064 (for J/Psi 3.096 GeV)
6.234 (for Psi' 3.686 GeV)
6.571 (for Psi 3.770 GeV)
6.991 (for X 3.872 GeV)
7.277 (for Y 3.940 GeV)
7.705 (for Psi 4.040 GeV)
8.685 (for Y 4.260 GeV)
11.917 (for D*_sJ D_s 4.9178 GeV)
15.000 (for Drell-Yan)
I propose 10,000,000 events for each.
Quote: | Could you, Soeren and Johan, please try to make a concise list of possible simulations to run, based on the packages which will be available for the DC01?
|
Paul, can you specify your question a bit?
Because all the mentioned above is the list.
1. tpc sim and reco (see macros, nEvents, pT above)
2. stt sim and reco (see macros, nEvents, pT above)
3. dpm event generation
(bash scripts by Johan, see p_beam, nEvents just a few lines above)
4. UrQMD event generation
(see separate posting by Johan)
This is it.
Quote: | * Storage:
One event is 250 Bytes (the number was given by Soeren). Is this true for any of the prposed simulations (UrQMD?)? If yes then we most probably have far enough storage capacity (1TB ~ 4.E9 events).
|
Paul, please be careful!!!
for stt and tpc sim and reco, it is much more!!!
It is the numbers in my last posting.
for tpc 56.5 kB=kiloBytes per event
for stt 17.3 kB=kiloBytes per event
plus an offset for each run (the "zero" events case).
cheers,
Soeren
[Updated on: Fri, 11 April 2008 15:40] Report message to a moderator
|
|
|
|
if you can wait, please wait a few hours ... [message #6364 is a reply to message #6361] |
Fri, 11 April 2008 12:05 |
Jens Sören Lange
Messages: 193 Registered: June 2005
|
first-grade participant |
From: *physik.uni-giessen.de
|
|
Hi everyone,
just to let everyone know ...
I know we fixed the PandaRoot revision, but if there is a chance to wait for a few hours, please do so ...
Mohammad and Stefano are working on some more things.
The idea is to switch on a few more detectors (incl. EMC reco -> clusters for later track matching for both tpc and stt) and Mohammad is just writing a interface to call dpm from a root macro (so that the bash scripts are not needed). We also discussed to run dpm full sim (Geant3, Geant4) instead of only dpm event generation.
So, -> a few nice things!
Anyway, if there is any chance that you wait for a few hours, this would be great ...
cheers, Soeren
|
|
|
|
final PandaRoot for DC [message #6378 is a reply to message #6370] |
Fri, 11 April 2008 15:50 |
Jens Sören Lange
Messages: 193 Registered: June 2005
|
first-grade participant |
From: *physik.uni-giessen.de
|
|
Hi all,
we proudly present
PandaRoot rev2480
-> this will now be the version for the DC.
/macro/dc1/stt
new: contains emc digis (clusters), too
for track matching
note: there is no stt rundigi.C macro anymore
-> it is not integrated into the runreco.C macro
(as it is also for tpc)
/macro/dc1/tpc
new: contains emc digis (clusters), too
for track matching
track visualization is now switched off
-> saves disk space
/macro/dc1/full
THIS IS NOW THE DPM !
(it means, basically no basf script necessary anymore,
but if you want to run the DPMGen executable for some
reason, this also still works).
new: we do not only event generation but even Geant3.
all detectors.
pT and nEvents are now parameters of the macro.
disk space: 105 kB per 1 event
(so not anymore only 250 bytes/event!)
CPU time: ~1 sec per 1 event.
note: there is no option for the seed.
Mohammad says that it is taken from the date.
NOTE! pgenerators/DpmEvtGen needs to be compiled on each site!
(this is not part of the cbuild)
but Florian will prepare it for the packages.
all is tested and works.
additional note:
all macros contain new newest mvd and newest forward emc geometry!
(they were generated with different external packages,
but Stefano fixed it).
Thanks everyone for debugging etc. (Mohammad, Stefano, Ola, Tobias, Florian, ...)
I think the phone companies will be very happy about today's accounting bill.
[Updated on: Fri, 11 April 2008 15:51] Report message to a moderator
|
|
|
|
|
|
Re: random number generator for DPMgen [message #6389 is a reply to message #6388] |
Fri, 11 April 2008 22:20 |
Johan Messchendorp
Messages: 693 Registered: April 2007 Location: University of Groningen
|
first-grade participant |
From: *xs4all.nl
|
|
Hi Mohammad and all others,
Thanks for your clear answer, Mohammad. I just wonder what the probability is that two (or more) jobs will start a DPM event with the same seed. The "time(NULL)" instruction gives an output in precision of seconds, right? Hmmm, in that case it could actually be very likely. It might, therefore, be more safe to set it up it via gRandom, which one can more easily control from the "outside". Actually, is this the same for the box generator (as used for the other DC macros)?
Best wishes,
Johan
(ps, really nice work what has been done for this DC. Thanks to everyone involved, really impressive!!!!!)
|
|
|
|
|
|
|
|
Run list for DC1 open for discussion [message #6471 is a reply to message #6431] |
Wed, 16 April 2008 16:51 |
Paul Buehler
Messages: 11 Registered: October 2007
|
occasional visitor |
From: *smi.oeaw.ac.at
|
|
Dear all,
I put a list of jobs to run during the DC on the wiki http://panda-wiki.gsi.de/cgi-bin/view/Computing/DataChallenge1 and would like to ask you for your comments and suggestions.
The list contains all the proposed scripts and momenta. However it foresees to produce only 100'000 events per momentum and not the 10'000'000 as originally proposed.
There are:
- 240 stt splits a 10'000 events, each running for ~2h, producing 200 GB in total
- 240 tpc splits, 10'000, ~2h, ~200 GB
- 110 full splits, 10'000, ~9h, ~200 GB
- 100 urqmd splits, 10'000, ~1h, ~1 GB
- 1000 generic splits, -, ~10 sec, 10 MB
In total: ~2000h, 1 TB
The critical point I think will be the available CPU time and data transfer time.
Alternatively we could run only a few momenta with higher statistics.
Paul
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Re: DC1 finished [message #6551 is a reply to message #6550] |
Mon, 21 April 2008 13:33 |
|
Dear all,
We are still updating the DC1 wiki. I will add final plots and graphs from MonaLisa shortly. Please watch this for updates.
|
|
|
|
Re: DC1 Summary [message #6557 is a reply to message #6553] |
Tue, 22 April 2008 08:09 |
Kilian Schwarz
Messages: 91 Registered: June 2004 Location: GSI, Darmstadt
|
continuous participant |
From: *gsi.de
|
|
great plots and description. Very informative.
I think the whole exercise was a big success !!!
Cheers,
Kilian
|
|
|
Goto Forum:
Current Time: Tue Dec 03 11:54:19 CET 2024
Total time taken to generate the page: 0.01091 seconds
|