[gmx-users] mdrun and simulation time

[Justin Lemkul]

On 5/13/13 6:41 AM, Francesco wrote:
> Good morning all,
> This morning I checked the output of a 8ns (4 x 2ns) of simulation and I
> noticed a strange behaviour:
> The fist two simulations (each 2ns) ended up correctly and they both
> took 2h 06min to finish.
> The second two were still running when the cluster time was over  (I
> asked for 2.30)  and
> were truncated.
> My system contains around 160k atoms and all the previous 2ns simulation
> took between 2h and 2h:10min (77 cores, no gpu).
>
> I had a look at the log and it seems that in the last two simulations
> mdrun
> did only 120.000 steps instead of 1.000.000.
>
> Is it strange or it is possible/normal that with the increase of the ns
> (always splitted in 2ns and extended) the running time is bigger?
>

Random performance loss often happens when one or more nodes being used for the 
job get stuck or have errors.  If you're submitting to a queuing system, there 
should be diagnostic information that your admins can access that would suggest 
why this is happening.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================