[Pw_forum] I/O performance on BG/P systems

[Gabriele Sclauzero]

Il giorno 26/mar/2011, alle ore 09.12, Paolo Giannozzi ha scritto:

> 
> On Mar 25, 2011, at 13:58 , Gabriele Sclauzero wrote:
> 
>> Before moving some complaints to the sysadmins, I would like to have
>> some feedback from those of you who have used this kind of  
>> architecture.
>> Are you aware of similar kind of I/O issues? Is there any performance
>> tuning that can be done on the software/compiler side?
> 
> I also noticed that reading/writing the CP restart file on a BG takes a
> sizable amount of time, in the order of a few minutes for a typical  
> system
> suitable for BG. I had attributed this to the "gathering" of the  
> distributed
> pieces of wavefunctions that are subsequently written to file by a  
> single
> process, though;

Are you referring to the wf_collect feature? I've not been using that, so I was supposing that each process writes independently its piece of wavefunctions (at least for PWscf), am I right?

> in your case, it is the I/O by each process that  
> takes a
> lot of time.

Do you think that this might be a symptom of a GPFS problem or misconfiguration? Have you ever observed such a behavior?

> You may want to try to reduce the I/O to the strict  
> minimum:
> work with one k-point per processor (use k-point parallelization if  
> needed),

Indeed that is what I am trying to do since I got aware of this problem, but it is sometimes painful because I am working with hexagonal cells and I sometimes end up with 6 k-points or, even worse, 9 or 15... moreover in BGP you can't reserve an arbitrary number of processors or nodes, but just a sub-partition of the machine, as you surely know. So I have sometimes to put some of the processors at rest.


> use disk_io='low' (or 'none'): the code should write only at the end  
> (or never).

I've tried to used disk_io='low' in some relax calculations, this might be a good compromise. However the code writes the wavefunctions only once at the end of the run. Wouldn't it be more reasonable to write them at the end of each ionic step? (just a personal opinion...)



Thanks,

GS


> 
> P.
> ---
> Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
> 
> 
> 
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum


§ Gabriele Sclauzero, EPFL SB ITP CSEA
   PH H2 462, Station 3, CH-1015 Lausanne

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110329/869b9bac/attachment.html>