mpiexec-0.4 (mpich-pbs integration)

Douglas Johnson djohnson at osc.edu
Thu Sep 21 06:00:07 PDT 2000


Hi,
	There is a new version of mpiexec available at

http://www.osc.edu/~djohnson/mpiexec

This program is intended to be a replacement for the mpirun scripts used
with the ch_gm device for Myrinet. Mpiexec uses the task manager library
of PBS to spawn copies of the executable on all the nodes in a PBS
allocation.

The goals of the program are as follows,

-Integration of parallel jobs into PBS. 
-Proper accounting and enforcement of cpu time and resource use for
parallel jobs (all processes are children of pbs_mom, of course there
are ways around it).
-More scalable, reliable parallel job startup for clusters. *

There are limitations the program. The tm layer is not reliable enough for
large processor count parallel jobs. There's a patch included against
pbs-2.2p11. It is possible to modify the program for use with other MPICH
devices.

This program is distributed under the GPL, see the LICENSE file included
with the program.

-- 
Doug Johnson
Systems Developer/Engineer
The Ohio Supercomputer Center
djohnson at osc.edu 








More information about the Beowulf mailing list