Contents
Download Meso-NH in binary standalone mode for DRIHM GRID
beta binary version of MNH 4-9-3 DRIHM_V0
Download the tar-ball MNH_DRIHM_V0.tar.gz file ( 180 MB ) , and untar it where you want
tar xvfz MNH_DRIHM_V0.tar.gz
Test Case
Without physiographic :: 002_3Drelief
Download the tar-ball 002_3Drelief test case file ( 5.3KB ) , and untar it where you want
tar xvfz PACK-002_3Drelief.tar.gz
download the test.sh file ,and run it
in this the ROOT_MESONH variable must point to the directory where the MNH binaries MNH_DRIHM_V0 was untarred
... # # ROOT_MESONH = where MNH binaries MNH_DRIHM_VO was untared # #export ROOT_MESONH=~/bin/MNH_DRIHM_V0 export ROOT_MESONH=`pwd`/MNH_DRIHM_V0 ...
With physiographic :: 007_16janvier
Download the tar-ball 007_16janvier test case file ( 23MB ) , and untar it where you want
tar xvfz PACK-007_16janvier.tar.gz
download the test.sh file ,and run it
In this test case, the first time, the 001_pgd1/get_pgd_files script will fetch via wget the physiographic files = 5 GB untarred, in the ~/PREP_PGD_FILES_WWW directory .
This script wait for an answer from the terminal for cancellation crt-C if not enough space is available . Must by adapted if no interactive session is done .
For example the and of test.sh could be rewrite like this
# # run the testcase # ... ./run_007_16janvier < /dev/null
Real MesoNH Drihm Case0
download the mesonh_real_drihm0.sh file , (rename it to test.sh if needed ) and run it .
Data will be fetch via wget from our MesoNH WEB site <=> 3GB for static PGD files + 1.5 GB for GRIB Time dependent files .
bandwidth problem now fixed about 10MB/sec expected , tell me (Juan) if not the case ...
The PREP_REAL_CASE exec which generate the MesoNH input files, will take about 4GB of Memory by MPI processes on this grid size, and if mpirun/m(va)pich is detected on the compute node the executable will be run in parallel with 2 MPI processes ( so 4GB*2 = 8GB needed ) . If needed change the MONORUN variable , to run with least memory = 1 MPI processes ( or more processes to run faster if more memory is available )
... #export MONORUN="Mpirun -np 2 " export MONORUN="Mpirun -np 1 " ...
Real MesoNH Drihm Case0 with datarsync-user acces
download the mesonh_real_drihm0_datarsync.sh file , (rename it to test.sh if needed ) and run it .
Data will be fetch via rsync port 2222 from our MesoNH WEB site <=> 3GB for static PGD files + 1.5 GB for GRIB Time dependent files .
bandwidth problem now fixed about 10MB/sec expected , tell me (Juan) if not the case ...
The PREP_REAL_CASE exec which generate the MesoNH input files, will take about 4GB of Memory by MPI processes on this grid size, and if mpirun/m(va)pich is detected on the compute node the executable will be run in parallel with 2 MPI processes ( so 4GB*2 = 8GB needed ) . If needed change the MONORUN variable , to run with least memory = 1 MPI processes ( or more processes to run faster if more memory is available )
... #export MONORUN="Mpirun -np 2 " export MONORUN="Mpirun -np 1 " ...