Vijay.M,
Many factors are involved in the speed of any calculation.
Are you using a single disk drive? The disk seeking that must occur to
write output to different locations on the disk may be causing I/O latency.
This is somewhat dependent upon the caching and buffering schemes used by
your system, but increasing the interval (number of steps) between which
information is written may help.
Also, if you have very large systems that, together, exceed available
memory, then some information may be written to disk in a "virtual memory"
scheme. Each time one of the calculations needs some information it may
need to load it from disk, a phenomenon called "thrashing" that can greatly
hurt performance.
You may wish to monitor the utilization of different resources during the
calculations to get a better idea of what the limiting factor is.
James
On Mon, Jul 12, 2010 at 4:54 AM, Vijay Manickam Achari <vjrajamany.yahoo.com
> wrote:
> Dear AMBER users,
>
> I am using SGI4700 to run my simulation. It contains 128 CPUs.
> I use 64 CPUs to run my job at per submission. If I submit two jobs, each
> one with 64 CPUs I notice my simulation runs very very slow.
>
> If I run single job with only 64 CPUs it runs fine. (THE OUT PUT is MOVE
> FASTER).
>
> I am not sure about why it run very slow if two jobs run and runs fine if
> there only one job running.
>
> Even though I reduce the CPUs become 32 at each submission, I notice the
> same problem occurs. I can not submit two jobs with 32 CPUs each. I only can
> submit single job with 32 CPUs.
>
>
> Could I get any help related to this issue?
> Is there any idea on how to solve this problem?
>
>
> Advance thanks for your help.
>
> Regards
> Vijay.M
>
>
>
>
>
>
>
>
>
>
> Vijay Manickam Achari
> (Phd Student c/o Prof Rauzah Hashim)
> Chemistry Department,
> University of Malaya,
> Malaysia
> vjramana.gmail.com
>
>
>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jul 12 2010 - 08:30:03 PDT
