Difference between revisions of "Building CESM models on the NSC triolith system"

From SNIC Documentation
Jump to: navigation, search
Line 21: Line 21:
Ensure the correct intel compile module and wrappers are loaded using:
Ensure the correct intel compile module and wrappers are loaded using:
:<code>module load build-environment/nsc-recommended</code>
:<code>module load build-environment/nsc-recommended</code>

Revision as of 14:04, 8 November 2013

This page describes the building the CESM series of models, including CCSM4 (e.g. CCSM4.0, CESM1.1, CESM1.2) on NSC's triolith system (triolith). Triolith consists of 1200 HP SL230s compute nodes is equipped with two Intel E5-2660 (2.2 GHz Sandybridge) processors with 8 cores each, i.e 16 cores per node. Triolith is equipped with a fast interconnect (Infiniband from Mellanox (FDR IB, 56 Gb/s) in a 2:1 blocking configuration) for high performance for parallel applications.

Triolith is a SNIC-funded system, and computing time on Triolith is exclusively allocated to SNIC projects (Large, Medium and Small). See [1] and [2] for information on how to apply for a SNIC project. Contact: support@nsc.liu.se.

CESM1 unix group on triolith

If you are going to run one of the CESM models on triolith, you should join the 'cesm1' unix group on triolith. This is not absolutely necessary but being a member allows access to shared storage space on /nobackup/global/proj/cesm1 where pre-downloaded initial data and ancillary files are stored. Email support@nsc.liu.se to request membership to the cesm1 unix group. Please note: Files that you copy to the shared /nobackup/global/proj/cesm1 will be charged to your space allocation on triolith.

The CESM models (incl. CCSM4) - Getting the source code

CESM is a free, open source model, but a short registeration is required before access to the subversion repository is granted. The source code, build and run scripts can be downloaded as described in the CESM1 user guide. To list the available model versions, after CESM registration (on triloith):

svn list https://svn-ccsm-release.cgd.ucar.edu/model_versions

The source files should be downloaded into the /home file system. This ensures that your code, and configuration files are backed up.

Building and configuring the model

Ensure the correct intel compile module and wrappers are loaded using:

module load build-environment/nsc-recommended

As, an example. to create a CESM1.1.1 version in /home/$USER/Projects/cesm1.1.1:

1. Change to the appropriate folder

cd Projects

2. Checkout version 1.1.1

svn co https://svn-ccsm-release.cgd.ucar.edu/model_versions/cesm1_1_1 cesm1.1.1

3. Files containing triolith specific machine environment variables have been created and are stored in the folder /nobackup/global/proj/cesm1/triolith_config_cesm1.1.1 (triolith_config_ccsm4, triolith_config_cesm1.0.4 and triolith_config_cesm1.2.0 config files also available). These files should be copied to the appropriate location in your local distribution. Based on the steps above:

cd /home/$USER/Projects/cesm1.1.1/scripts/ccsm_utils/Machines
cp /nobackup/global/proj/cesm1/triolith_config_cesm1.1.1/* .

Note: for CCSM4.0, the file ncdio.F90, also found in the /nobackup/global/proj/cesm1/triolith_config_ccsm4 must be copied to the ${CCSM4 base dir}/models/lnd/clm/src/main/ folder. This file corrects a small bug compile error in the land model.

4. CESM1 can now be built in the usual way as described in the CESM1 users guide (see Creating and Setting Up A Case) using 'triolith' for the machine designation. For example, to create a new 'f19_g16' resolution case using the 'B' compset:

cd /home/$USER/Projects/cesm1.1.1/scripts/
./create_newcase -case ../cases/lbtest-f19_g16-B -res f19_g16 -compset B -mach triolith -pecount M

5. Configure and build your model:

cd /home/$USER/Projects/cesm1.1.1/cases/lbtest-f19_g16-B

Note: for CCSM4 CESM1.0.X and NorESM use

cd /home/$USER/Projects/cssm4.0/cases/lbtest-f19_g16-B
./configure -case

6. Edit the env_run.xml file for the particular simulation you wish to make, then submit your job:

sbatch ./lbtest-f19_g16-B.run

Model load balancing

Load balancing of the models (laying out pe configurations e.g. load balancing a case) has been completed for a limited set of model configurations (resolution f19_g16, compset B). Using the 'pecount' keyword when invoking .create_newcase (see 4. above) will create a load balanced configuration of the model (res f19_g16, compest B). pecount options are 'S' (small), 'M' (medium) and 'L' (large): see EC-Earth and CESM model timings. For assistance with load balancing alternative configurations (resolutions / compsets), email: support@nsc.liu.se.