WRF computational requeriments

[mamartin]

Hello! I would like to know what computational requirements (number of processors, speed or other considerations of interest) would you recommend me for generating hourly meteorological data for 1 year to use in dispersion models. Currently it takes me 4-5 hours to generate 1 day, so the time to run 1 year is very high. My smallest domain is 100*100 points and 31 levels high.

On the other hand, I think that using MPI improves times, but in the tests that I have carried out, I have not obtained improvements when going from 4 to 12 processors. For use MPI, I execute "mpirun -np 4 wrf.exe" (or -np 12). I need to do something else?

Thank so much for your help.

 

[kwerner]

Hi,
There are many things to consider regarding the right set-up to use, and many different things can make the simulation take longer - depending on various physics options, the actual system, size and resolution of domains, etc. It's difficult for us to tell you what would be best. However, I can point you to the FAQ that discusses how to determine a good number of processors for your case, which may be a good starting place.

 

[mamartin]

ok, thank you very much.

I attach my namelist, wps and input, in case someone can tell me something that is slowing down the run.

I really appreciate your help, since the times I need must be much lower and I am looking for solutions.

 

[Ming Chen]

Please take a look at the website here, which describes in detail computer hardware/software requirements:
https://forum.mmm.ucar.edu/phpBB3/viewtopic.php?f=71&t=70

 

[Ming Chen]

unfortunately we don't have a software engineer to provide specific recommendations based on your case. Hope someone in the community may provide more information.