[gmx-users] Using Gromacs on Condor

[Mark Abraham]

On 4/27/2011 9:08 PM, Natalie Stephenson wrote:
> Hi,
> I was wondering if anyone new a way to run mdrun commands through a 
> condor system.  I've been told by the people that run our system that 
> unless I can split it into seperate threads then I would not be able 
> to run it as a job, because it would just get stopped in the morning - 
> which from what I've seen would just end my simulation and I would 
> have to start again.

Condor is not a suitable platform for running molecular dynamics in 
parallel, unfortunately. Low-latency communication is required if the 
processors do not share the same piece of silicon, and nobody running 
Condor will have that. You can get some value from Condor if you have 
several simulations to run, by making it easy to run each one on a 
different machine.

Starting and stopping simulations is not much of a problem, with 
checkpointing in GROMACS 4.x.
> Currently my simulations are taking around a month - which as I'm 
> nearing the end of this project is getting quite frustrating as you 
> can imagine.  Any ideas on how I can speed this without buying a new 
> computer (which is an idea I've had veeto'd from my supervisor many 
> many times!) would be much appreciated!!!

Sounds like you need to work with your supervisor to change 
expectations, or look around for some free time you can ask for... try 
big or new computing facilities, especially new ones with 
recently-installed under-utilized hardware.

Mark



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110427/dd3241ca/attachment.html>