Hi,
On Fri, Jan 8, 2021 at 11:08 AM Maghsoud, Yazdan
<YazdanMaghsoud.my.unt.edu> wrote:
>
> Dear Amber Users,
>
> I'm doing clustering for a system of around 150,000 atoms (water molecules are also included) for further QM/MM calculations. I'm doing this clustering for three replicates each 500ns and a total of 150 .mdcrd files (each mdcrd file contains 2500 frames). I'm also using 20 CPUs and 120GB of RAM but after 20% of processing I face the following error:
That is a *lot* of data. 150000 atoms * 3 is 450k coordinates per
frame, times 150 trajectories of 2500 frames means you will need about
675 GB to store that trajectory in single precision in memory (4 bytes
per coord); note that 20% of that is just over your system memory plus
whatever caching, which is probably why you're crashing. You have two
ways around this:
1) Strip away the water if you don't need to cluster on it. That
should reduce your system size significantly. If you plan on doing a
lot of calculations that don't involve the water, consider making a
separate "stripped" trajectory as this will speed up calculations
quite a bit. Make sure you save the equivalent stripped topology (via
'strip parmout', 'strip outprefix', 'parmstrip', or use parmed).
2) Load the trajectories as on-disk data sets (TRAJ data sets in
cpptraj terms). You won't be able to manipulate them but they will
stay on disk.
3) Cluster on only a subset of the trajectory frames. You can do this
multiple times for different subsets and compare the results. You
should probably be doing this anyway in order to test whether your
clustering parameters are appropriate for your system.
With a lot of frames you might also run into the problem that your
pairwise distance matrix is too big to fit in memory. In that case
you'll have to use 'pairwisecache disk' or 'pairwisecache none'; be
aware that both are slower than having the pairwise matrix in memory.
Hope this helps,
-Dan
>
> = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
> = PID 94219 RUNNING AT c11-8-4
> = EXIT CODE: 9
> = CLEANING UP REMAINING PROCESSES
> = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
> ===================================================================================
> YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Killed (signal 9)
> This typically refers to a problem with your application.
>
>
> Thanks for any help in advance,
>
> Yazdan
> _______________________________________________
> 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 Fri Jan 08 2021 - 10:00:02 PST
