User Tools

Site Tools


hpc:bghep:benchmarks

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
hpc:bghep:benchmarks [2014/02/17 19:31]
asc [Fig 3]
hpc:bghep:benchmarks [2014/05/27 16:45] (current)
edmay
Line 48: Line 48:
 R == sec/event  =       job_run_time/total_number_of_events R == sec/event  =       job_run_time/total_number_of_events
 Nc == number of cores used in job Nc == number of cores used in job
-<code>+</code>
  
 For a perfect speed-up this would always be 1. My experience with clusters of For a perfect speed-up this would always be 1. My experience with clusters of
Line 61: Line 61:
 {{:hpc:bghep:alcf-vesta-pythia6-promc-mpi-d.png?400|}} {{:hpc:bghep:alcf-vesta-pythia6-promc-mpi-d.png?400|}}
  
 +The ALCF experts suggested the I/O model of 1 directory and many files in that 1 directory would preform badly due to lock contention
 +on the directory! Thus the example code was modified to use a model
 +of 1 output promc data file per directory. Running the modified code
 +produced the following figures:
 +http://www.hep.anl.gov/may/ALCF/new.vesta.4up.pdf Vesta Plots
 +Focusing on the 'Efficiency' plots there appears to be some (small)
 +improvements both at low core numbers and a high core numbers: 80% rising to 90% and 20% rising to 40% respectively.
 +
 +The large step between 512 to 1024 is still present!
 +
 +As part of the bootcamp for MIRA the code was moved to the
 +BG/Q Mira and a subset up the benchmarks were run in the new 
 +IO model. The results are shown in 
 +http://www.hep.anl.gov/may/ALCF/new.mira.4up.pdf Mira figures
 +Again focusing on the Efficiency plot the results are very
 +similar. This suggest when using this naive IO model of each
 +mpi rank writing its own output ProMC file should be limited to
 +jobs of 512 cores or less for good utilization of the machine
 +and IO resouces. This is OK for Vesta where the minimum charging
 +is 32 nodes (ie 512 cores). While on Mira the minimum is 512 
 +nodes (ie 8192 cores) there is not a good match!
  
 [[hpc:bghep:| << back]] [[hpc:bghep:| << back]]
  
- --- //[[[email protected]|Ed May]] 2014/02/04 10:42//+ --- //[[[email protected]|Ed May]] 2014/02/04 // 
 + --- //[[[email protected]|Ed May]] 2014/05/27 // 
  
  --- //[[[email protected]|Sergei Chekanov]] 2014/02/04 10:42//  --- //[[[email protected]|Sergei Chekanov]] 2014/02/04 10:42//
hpc/bghep/benchmarks.1392665460.txt.gz · Last modified: 2014/02/17 19:31 by asc