[Wrf-users] How many CPU I should allocate for 2 year hind-cast
agnes.mika at bmtargoss.com
Tue Mar 16 02:51:14 MDT 2010
mmkamal at sciborg.uwaterloo.ca wrote:
> I would like to run a two year hind-cast simulation for the period of
> 2007-2008 over North-Eastern North America with 3 nest. The
> namelist.wps file is given below.
> Would you please tell me how many CPU I should allocate to finish the
> run successfully with limited duration (roughly ~ 15 days). I am using
> WRF 3.1.1.
I am not familiar with your architecture so can't give you a detailed
advice. What we normally do when we have to estimate how long a
hindcast would take to compute is to do some timing tests. Go,
experiment a bit with the number of CPU's and see what works best.
WRF doesn't scale too well, so above a certain number of processors
you are better off with starting too forecasts in parallel (e.g. 2007
on half of the processors and 2008 on the other) then throwing all
your processors on one run.
A tip for reducing the time you need to complete your hindcast: in our
case the most time is consumed by I/O (writing the wrfout and wrfrst
files). You could set up the registry so that the wrfout files will
only contain the variables you want to keep (I wouldn't touch the
wrfrst files) and then re-compile the model. This way you can reduce
the time needed for I/O considerably.
Dr. A'gnes Mika
Advisor, Meteorology and Air Quality
Tel: +31 (0)527-242299
Fax: +31 (0)527-242016
P.O. Box 61, 8325 ZH Vollenhove
Voorsterweg 28, 8316 PT Marknesse
Confidentiality Notice & Disclaimer
The contents of this e-mail and any attachments are intended for the
use of the mail addressee(s) shown. If you are not that person, you
are not allowed to take any action based upon it or to copy it,
forward, distribute or disclose its contents and you should delete it
from your system. BMT ARGOSS does not accept liability for any errors
or omissions in the context of this e-mail or its attachments which
arise as a result of internet transmission, nor accept liability for
statements which are those of the author and clearly not made on
behalf of BMT ARGOSS.
More information about the Wrf-users