[Wrf-users] How many CPU I should allocate for 2 year hind-cast

[Andrew Porter]

Hi Kamal,

Your best bet is to do some small experiments to see what your optimum 
job size is. Given your domain sizes I'd guess it's in the region of 512 
MPI processes so try running a short version of your job (maybe just 30 
minutes of simulated time) on 128, 256 and 512 processes.

You can look at the time taken to step the model (including nested 
grids) by grep'ing rsl.out.0000 for the lines like:

Timing for main: time 2008-07-31_00:01:48 on domain   1:    8.95818 
elapsed seconds.

(this method excludes the start-up time which, for a short job, can be a 
significant %).

HTH.

Cheers,

Andy.

>
> 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.
>
>
>
> &share
> wrf_core = 'ARW',
> max_dom = 3,
> start_date =
> '2007-01-01_00:00:00','2007-01-01_00:00:00','2007-01-01_00:00:00',
> end_date = '2008-12-31_00:00:00','2008-12-31_00:00:00','2008-12-31_00:00:00',
> interval_seconds = 10800,
> io_form_geogrid = 2,
> /
>
> &geogrid
> parent_id = 0, 1, 2,
> parent_grid_ratio = 1, 3, 3,
> i_parent_start = 1, 124, 91,
> j_parent_start = 1, 106, 42,
> s_we = 1, 1, 1,
> e_we = 325, 181, 154,
> s_sn = 1, 1, 1,
> e_sn = 250, 163, 202,
> geog_data_res = '10m','5m','2m',
> dx = 8000,
> dy = 8000,
> map_proj = 'lambert',
> ref_lat = 43.,
> ref_lon = -80.,
> truelat1 = 30.,
> truelat2 = 60.,
> stand_lon = -80,
> geog_data_path = '/work/kamal/DATA/geog'
> /
> &ungrib
> out_format = 'WPS',
> prefix = 'NARR_FLX',
> /
>
> &metgrid
> fg_name = 'NARR'
> io_form_metgrid = 2,
> constants_name ='./NARRFIX:1979-11-08_00'
> /
>
> &mod_levs
> press_pa = 201300 , 200100 , 100000 ,
> 95000 , 90000 ,
> 85000 , 80000 ,
> 75000 , 70000 ,
> 65000 , 60000 ,
> 55000 , 50000 ,
> 45000 , 40000 ,
> 35000 , 30000 ,
> 25000 , 20000 ,
> 15000 , 10000 ,
> 5000 , 1000
> /
>
>
> ========================================================================
> The system configuration where I am going to run WRF is given below:
> ========================================================================
> Manufacturer HP
> Operating System XC 4.0 (RHEL 5.1)
> Interconnect InfiniBand
> Total processors/cores 2704
>
> Nodes 0-0 8 cores
> 2 sockets x 4 cores per socket
> Xeon @ 2.83 GHz
> Type: Admin
> Notes: Admin node
> Memory: 16.0 GB
> Local storage: GB
>
>
> 1-336 8 cores
> 2 sockets x 4 cores per socket
> Xeon @ 2.83 GHz
> Type: Compute
> Notes: Compute nodes
> Memory: 16.0 GB
> Local storage: GB
>
>
> 377-377 8 cores
> 2 sockets x 4 cores per socket
> Xeon @ 2.83 GHz
> Type: Login
> Notes: Login node
> Memory: 16.0 GB
> Local storage: GB
>
> Total attached storage 130000.0 GB
> Suitable use Parallel applications
>
>
>
>
> 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.
>
>
>
>
> Thanks in advance
> Kamal
> University of Waterloo
> Canada
>
>
>
>
>
> _______________________________________________
> Wrf-users mailing list
> Wrf-users at ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/wrf-users


-- 
Dr. Andrew Porter
Computational Scientist
Advanced Research Computing Group
Computational Science and Engineering Dept.
STFC Daresbury Laboratory
Keckwick Lane
Daresbury
WA4 4AD
Tel. : +44 (0)1925 603607
email: andrew.porter at stfc.ac.uk