Hi,
On Mon, May 25, 2015 at 4:08 PM Zhang, Cheng c.zhang...@ucl.ac.uk wrote:
Dear GROMACS,
I am now using openmpi nodes to run GROMACS (e.g. mdrun_mpi) on our
cluster. When the nodes required are too many (e.g. more than 8), jobs
always take a long time to wait in the queue. So I wonder if there is a
possibility that we can
1) convert the job into many serial jobs?
Only if you are prepared to run a separate simulation in every concurrent
single-node job, or you can get better throughput by chaining single-node
jobs than you can get by running in parallel. If you have measured this,
you should complain bitterly - people probably paid good money for the
network, and you would have evidence that it got bought for the
pre-production benchmarks, and is now being wasted in practice. If they
wanted to build a single-node throughput cluster, then they should do that
and e.g. not bother to install MPI on it. ;-)
2) convert the job into several openmpi jobs but with only 1 node in each
job?
No. All the cores synchronize every few milliseconds. Nothing can start
until all the nodes are ready.
When serial jobs are submitted, hundreds of jobs can be run within just a
few minutes.
Sounds like you will be able to generate a good complaint :-)
Mark
Thank you very much.
Yours sincerely
Cheng
--
Gromacs Users mailing list
* Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
posting!
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
send a mail to gmx-users-requ...@gromacs.org.
--
Gromacs Users mailing list
* Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a
mail to gmx-users-requ...@gromacs.org.