Yeah I just ran IU's nightly set against my DB and it went fine submitting 6024 test_run results.
I say we are good to go forward.

Here are some performance numbers:
Summary Report (24 hours all orgs)
[ 107,070 test_runs]
v2 -  87 sec
v3 -   6 sec
Summary Report (24 hours Only 'iu')
[ 29,625 test_runs]
v2 -  37 sec
v3 -   4 sec

Summary Report (Past 3 days all orgs)
[ 294,395 test_runs]
v2 - 138 sec
v3 -   9 sec

Summary Report (Past 3 days Only 'iu')
[ 86,026 test_runs]
v2 -  49 sec
v3 -  11 sec

Summary Report (Past 2 weeks all orgs)
[ 1,460,824 test_runs]
v2 - 863 sec
v3 -  34 sec
Summary Report (Past 2 weeks Only 'iu')
[ 346,804 test_runs]
v2 - 878 sec (not seeded)
v2 -   2 sec (pre-seeded)
v3 -  12 sec


Summary Report (Past 1 month all orgs)
[2,981,678 test_runs]
v2 - 1395 sec
v3 -  158 sec
Summary Report (Past 1 month Only 'iu')
[ test_runs]
v2 - 1069 sec (not seeded)
v2 -    2 sec (pre-seeded)
v3 -   39 sec

Summary Report (2007-06-18 - 2007-06-19 all orgs)
[ 43,816 test_runs]
v2 - 484 sec
v3 -   5 sec
Summary Report (2007-06-18 - 2007-06-19 only 'iu')
[ 30,059 test_runs]
v2 - 479 sec (not seeded)
v2 -   2 sec (pre-seeded)
v3 -   2 sec


Some stats I collected that we need not distribute, but thought were interesting:
Some intersting stats to date (2007-08-20)
# test_run results       : 17,522,720
# test_build results     :     42,655
# mpi_install results    :     45,096
# of performance results :     58,365

Most popular test suites:
  intel    : 65.5 %
  ibm      : 28.0 %
  onesided :  4.3 %
  trivial  :  1.5 %
  imb      :  0.4 %

Most popular compilers:
  gnu       : 49.8 %
  intel     : 30.8 %
  sun       :  6.9 %
  pathscale :  6.8 %
  pgi       :  5.7 %

OS testing percentages:
  Linux: 93.2 %
  SunoS:  6.8 %

Org. Contribution percentages:
  Cisco    : 66.4 %
  IU       : 23.2 %
  Sun      :  6.8 %
  IBM      :  2.3 %
  HLRS     :  1.3 %
  Voltaire :  0.0 %
  UTK      :  0.0 %


On Aug 23, 2007, at 12:35 PM, Jeff Squyres wrote:

I think we got good submits to the josh db last night -- I think my
prior problems were pilot error (accidentally sharing scratch
directories between my josh-db testing MTT runs and my production MTT
runs).

http://osl.iu.edu/~jjhursey/research/mtt/server/php/reporter.php?
do_redir=6

So I think Cisco is happy with the new DB and good for the Monday
rollout plan.

--
Jeff Squyres
Cisco Systems

_______________________________________________
mtt-devel mailing list
mtt-de...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel

Reply via email to