Difference between revisions of "Comparison of CyberShake Studies"

From SCECpedia
Jump to navigationJump to search
Line 4: Line 4:
 
| [[CyberShake Study 15.12]] || December 3, 2015 - ???  
 
| [[CyberShake Study 15.12]] || December 3, 2015 - ???  
 
|-
 
|-
| [[CyberShake Study 15.4]] || April 16 - May 24, 2015 || 1 Hz CyberShake map, with AWP-ODC-GPU, G&P 2014, CVM-S4.26, and new sites. Calculated RotD also. || 1 Hz || G&P 2014 || CVM-S4.26 || 194 || 336 || NCSA Blue Waters, OLCF Titan (pilot jobs) || 37.6 million || 237,984 || 5505 || 4372 || 7.1 million || 7.7 TB || Used [[DirectSynth]] to reduce the number of post-processing tasks and cut I/O by 99.9%.
+
| [[CyberShake Study 15.4]] || April 16 - May 24, 2015 || 1 Hz CyberShake map, with AWP-ODC-GPU, G&P 2014, CVM-S4.26, and new sites. Calculated RotD also. || 1 Hz || G&P 2014 || CVM-S4.26 || 914 || 336 || NCSA Blue Waters, OLCF Titan (pilot jobs) || 37.6 million || 237,984 || 5505 || 4372 || 7.1 million || 7.7 TB || Used [[DirectSynth]] to reduce the number of post-processing tasks and cut I/O by 99.9%.
 
|-  
 
|-  
 
| [[CyberShake Study 14.2]] || February 18 - March 4, 2014 || 4 hazard models: CVM-S4.26 with CPU, CVM-S4.26 with GPU, 1D model with CPU, and CVM-H, no GTL with GPU.|| 0.5 Hz || G&P 2010 || CVM-S4.26, BBP 1D, CVM-H no GTL || 342 || 1144 || NCSA Blue Waters || 15.8 million || 295,040 || 99.9 million || 100707 (Condor), 29796 (submitted to Blue Waters) || 16 million || 57 TB || Workflow job to generate AWP workflow to support dynamic number of cores, GPU and CPU executables, batched SeisPSA jobs
 
| [[CyberShake Study 14.2]] || February 18 - March 4, 2014 || 4 hazard models: CVM-S4.26 with CPU, CVM-S4.26 with GPU, 1D model with CPU, and CVM-H, no GTL with GPU.|| 0.5 Hz || G&P 2010 || CVM-S4.26, BBP 1D, CVM-H no GTL || 342 || 1144 || NCSA Blue Waters || 15.8 million || 295,040 || 99.9 million || 100707 (Condor), 29796 (submitted to Blue Waters) || 16 million || 57 TB || Workflow job to generate AWP workflow to support dynamic number of cores, GPU and CPU executables, batched SeisPSA jobs

Revision as of 20:35, 13 February 2017

Study Name Date Science Goals Frequency Rupture Variations Velocity Model(s) Makespan (hrs) Number of sites Machine Core Hours Max Processors Tasks Jobs Files produced Data Comments
CyberShake Study 15.12 December 3, 2015 - ???
CyberShake Study 15.4 April 16 - May 24, 2015 1 Hz CyberShake map, with AWP-ODC-GPU, G&P 2014, CVM-S4.26, and new sites. Calculated RotD also. 1 Hz G&P 2014 CVM-S4.26 914 336 NCSA Blue Waters, OLCF Titan (pilot jobs) 37.6 million 237,984 5505 4372 7.1 million 7.7 TB Used DirectSynth to reduce the number of post-processing tasks and cut I/O by 99.9%.
CyberShake Study 14.2 February 18 - March 4, 2014 4 hazard models: CVM-S4.26 with CPU, CVM-S4.26 with GPU, 1D model with CPU, and CVM-H, no GTL with GPU. 0.5 Hz G&P 2010 CVM-S4.26, BBP 1D, CVM-H no GTL 342 1144 NCSA Blue Waters 15.8 million 295,040 99.9 million 100707 (Condor), 29796 (submitted to Blue Waters) 16 million 57 TB Workflow job to generate AWP workflow to support dynamic number of cores, GPU and CPU executables, batched SeisPSA jobs
CyberShake Study 13.4 April 17 - June 17, 2013 4 hazard models: CVM-S and CVM-H with the RWG V3.0.3 SGT code and AWP-ODC-SGT. 0.5 Hz G&P 2010 CVM-S, CVM-H 716 1132 NCSA Blue Waters (custom solution), TACC Stampede (Pegasus/Condor/Globus) 10.8 million (Blue Waters); 1.4 million (Stampede) 128,000 (Blue Waters); 17,600 (Stampede) 470 million (Stampede) 21912 submitted to Stampede 2264 (Blue Waters); 16 million (Stampede) 43 TB (Blue Waters), 12.3 TB (Stampede) Hierarchical DAXes with 8 sub-DAXes, PMC, pipe forwarding, Handoff jobs from Blue Waters to Stampede, merged SeisPSA jobs, in-memory rupture variation generation
CyberShake Study 2.2 (post-processing only) October 8, 2012 - March 6, 2013 (with a break from October 16 - November 8) Redid post-processing of CyberShake 1.0 with G&P 2010. 0.5 Hz G&P 2010 CVM-S 2976 217 NICS Kraken PMC, pipe forwarding
CyberShake 1.4 August 25, 2011 - March 14, 2012 Deterministic to 0.5 Hz + stochastic to 10 Hz, CVM-S4 and CVM-H 11.2 for PBR, SCSN, and near-SCSN sites, using G&P 2010 10 Hz stoch G&P 2010 CVM-S, CVM-H 4824 90 USC HPCC, TACC Ranger 112 million 112 million 4 TB Broadband
CyberShake 1.0 April 16 - June 10, 2009 0.5 Hz hazard map in Southern California, CVM-S 0.5 Hz G&P 2007 CVM-S 1314 223 TACC Ranger 6,900,000 14,544 192 million 3.9 million (Condor), 3952 (submitted to Ranger) 190 million 10.6 TB PDAX with 80 sub-DAXes, clustering, Condor glideins with Corral, zip files, Run Manager
CyberShake 0.5 (post-processing only) June 26 - July 18, 2008 Initial calculation of hazard curves 0.5 Hz G&P 2007 CVM-S 528 9 NCSA Mercury 442,400 800 7.5 million 22 (glideins) 7.5 million 90 GB PDAX with 80 sub-DAXes, clustering, Condor glideins, zip files