Download presentation
Presentation is loading. Please wait.
Published byKelley Osborne Modified over 9 years ago
1
www.egi.eu EGI-InSPIRE RI-261323 EGI-InSPIRE www.egi.eu EGI-InSPIRE RI-261323 VO auger experience with large scale simulations on the grid Jiří Chudoba (Institute of Physics and CESNET) with input from the Auger production team (J.Lozano Bahilo, G.Rubio, M.D.Serrano - UGR) and Jean-Noel Albert (LAL)
2
www.egi.eu EGI-InSPIRE RI-261323 PAO is an astroparticle project to measure ultra–high energy cosmic rays 12.4.2013 Jiri.Chudoba@cern.ch2
3
www.egi.eu EGI-InSPIRE RI-261323 12.4.2013 Jiri.Chudoba@cern.ch3
4
www.egi.eu EGI-InSPIRE RI-261323 VO auger for Worldwide Collaboration 12.4.2013 Jiri.Chudoba@cern.ch4 VO created in 2006 by the Prague group (CESNET and FZU) Now supported by 23 sites in 10 countries 63 members Main goal: simulations of cosmic ray showers using CORSIKA 23 sites 10 countries
5
www.egi.eu EGI-InSPIRE RI-261323 VO management Perun also manages accounts on UIs and fills VO members mailing list 12.4.2013 Jiri.Chudoba@cern.ch5
6
www.egi.eu EGI-InSPIRE RI-261323 JMM Production Job Management Module (JMM) written in bash and python automates all steps in the job management parameters to control behaviour (#jobs/collection, #jobs/submission cycle) status of jobs regularly checked 12.4.2013 Jiri.Chudoba@cern.ch6
7
www.egi.eu EGI-InSPIRE RI-261323 12.4.2013 Jiri.Chudoba@cern.ch7 Input sandbox Output sandbox (job information) Input cards generation Input cards generation Input cards Templates: JDL and script Templates: JDL and script DB DB gets job status and other relevant information from JMM PHP web server produces updated web pages Input cards associated to each shower of a library are generated Template JDL files and scripts are also modified for the given library JMM takes those inputs and submit jobs in an automated way Grid Computing Model
8
www.egi.eu EGI-InSPIRE RI-261323 Grid Computing Model 12.4.2013 Jiri.Chudoba@cern.ch8 Discard CEs: -Downtime -Ratio of aborted or failed jobs too high Discard CEs: -Downtime -Ratio of aborted or failed jobs too high - Maximum #jobs to submit - #Jobs per collection - Ratio running/waiting for new submission - …. - Maximum #jobs to submit - #Jobs per collection - Ratio running/waiting for new submission - …. Storage for: - Job status - CE, SE, CPU time, file sizes - …. Storage for: - Job status - CE, SE, CPU time, file sizes - ….
9
www.egi.eu EGI-InSPIRE RI-261323 Data Distribution 12.4.2013 Jiri.Chudoba@cern.ch9 160 TB 0.54 TB/day One “random” SE from predefined list is used to store output Close SE is favoured
10
www.egi.eu EGI-InSPIRE RI-261323 Data transfers to CC IN2P3 Decommissioning of an SE with many auger files 12.4.2013 Jiri.Chudoba@cern.ch10 FTS transfers from Lille to Lyon 2 months, 1.9 M files, 38.7 TB less than 1% of lost files 31 750 operations/day, 1300 ops/hour 650 GB/day, 27 GB/hour, 8 MB/s FTS transfers from Bordeaux to Lyon 1 month, 700 K files, 7.1 TB.6% of lost files 12200 operations/day, 500 ops/hour 160 GB/day, 7 GB/hour, 2 MB/s Many more small files in Bordeaux Large files stored to tapes in Lyon
11
www.egi.eu EGI-InSPIRE RI-261323 Usage 12.4.2013 Jiri.Chudoba@cern.ch11 Steep increase in last 3 years The VO auger dominates the usage in Astronomy and Astroparticle cluster in top ten since 2010 #5 (after LHC projects)
12
www.egi.eu EGI-InSPIRE RI-261323 Effectiveness evaluation Efficiency: cputime/walltime 12.4.2013 Jiri.Chudoba@cern.ch12
13
www.egi.eu EGI-InSPIRE RI-261323 Top ten VOs efficiency 12.4.2013 Jiri.Chudoba@cern.ch13 Efficiency of the biggest VOs for 2012-01 to 2012-12
14
www.egi.eu EGI-InSPIRE RI-261323 VO auger efficiency 12.4.2013 Jiri.Chudoba@cern.ch14 From 2012-01 to 2013-03 - efficiency improves
15
www.egi.eu EGI-InSPIRE RI-261323 Effectiveness evaluation Effectiveness = 12.4.2013 Jiri.Chudoba@cern.ch15 cputime of jobs with good output total walltime Difficult to estimate No information about cancelled or lost jobs Some jobs without job log file stored correct results Production maximizes throughput Just one of many possible definitions
16
www.egi.eu EGI-InSPIRE RI-261323 DIRAC for auger evaluation LFC future support was doubtful, DFC as an alternative Now: DPM consortium includes LFC into its portfolio LFC and DFC commands are not mapped one to one, our code requires changes DIRAC for job management pilot jobs may increase effectiveness First tests were done using the French NGI DIRAC instance (https://dirac.france-grilles.fr/DIRAC/)https://dirac.france-grilles.fr/DIRAC/ test jobs executed We applied for a small project to do larger scale tests 12.4.2013 Jiri.Chudoba@cern.ch16
17
www.egi.eu EGI-InSPIRE RI-261323 Instead of conclusions We thank all sites supporting the VO auger for their hardware resources and manpower support 12.4.2013 Jiri.Chudoba@cern.ch17 Achievements of the VO auger would be impossible without sites
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.