Sıra | DOSYA ADI | Format | Bağlantı |
---|---|---|---|
01. | Tobias Package Analysis Simulated | pptx | Sunumu İndir |
Transkript
1/24/20Mitglied der Helmholtz-GemeinschaftRadiation damage analysisfor the MVD on the pandaGRID| Tobias Stockmanns
1/24/20 Folie 2Tobias Stockmanns Analysis done on the PANDA GRIDData Challenge 02 (17.-21. November 2008)• Simulation of anti-proton on 132Xe with different beam momenta of 2.0 – 6.2 – 15.0 GeV/c• Full Panda detector geometry to include backscattered particles• Modification of MVD code to “see” neutral particlesBeam momentum[GeV/c]Simulated events2.0 645,0006.2 627.90015.0 487,000
1/24/20 Folie 3Tobias Stockmanns Analysis of the simulated data:• translate the flux of the different particle types into a normalized flux of 1 MeV neutronsAnalysis done on the PANDA GRID• Installation of a new package radDamAna::v1.2• this package is an updated version of pandaROOT• better would be to use a patch because the modifications on pandaRoot are very small• Problems:• Package does not compile/run on all sites• Errors: SE full / no cmake installed / wrong path variable• Additional problem: no error treatment
1/24/20 Folie 4Tobias Stockmanns Analysis done on the PANDA GRIDThe analysis is done with collections:• find -c <collectionName> <folder> <fileName> puts all files belonging to a simulation run into one collection• The files are still stored on the initial storage elements but the collection can be used as an input into a masterJob• The subJobs are distributed to the compute elements where the simulated file is stored• Advantage: No need to copy the large simulation files from one compute node to another• Problem: Two conditions for a subJob to run:• The necessary package has to be installed• The file to analyze has to be on the storage element close to the compute node• If one of them fails the data is not processed
1/24/20 Folie 5Tobias Stockmanns Analysis done on the PANDA GRIDProblems:• The installation of the package failed in Dubna (at the beginning) but it reported that the installation was successful Dubna took many subJobs but was not able to process them and they were lost for others• At GSI jobs start delayed (up to days) or sometimes not at all even if they meet all requirements• Once GSI is running it covers 90 % of the analyzed files (Dubna 7 %, Jülich 2.5 %) Beam momentum [GeV/c]Simulated eventsAnalyzed events on the GRIDEvents analyzed at desktop2.0 645,000 594,500 06.2 627.900 564,200 015.0 487,000 390,700 215,700
1/24/20 Folie 6Tobias Stockmanns Primary Particlesprimary particles:• neutrons, protons and pions• 33 primaries per event• almost spherical distribution ofmomentum15 GeV/c antiprotons on Xe
1/24/20 Folie 7Tobias Stockmanns Overview of the flux distributionAssumptions:• Interaction rate of 1x106• Operation time: one year with 50% duty cyclenew Class DrawEveHisto to plot histograms in EVE
1/24/20 Folie 8Tobias Stockmanns Pixel SDKSmall disks:• maximum flux for disks• 3 x 1010 neq[1MeV]/cm2• strong asymmetry
1/24/20 Folie 9Tobias Stockmanns Pixel BL1Pixel Barrel Layer 1:• highest particle flux• 4 x 1010 neq[1MeV]/cm2• strong asymmetry
1/24/20 Folie 10Tobias Stockmanns Summary• First real test of a data analysis running on pandaGRID was successful Thank you Dan, Kilian and Rene• Still a couple of problems have to be solved:• Availability of sites• Improvement of error tolerance• Improvement of documentation• Simplification/Improvement of tools on the GRID• Suggestions:• Automatic installation of pandaRoot every week/month• Use the GRID! It is very powerful and easier than it seems