Underworld

a long term geodynamics simulation platform

Underworld Community Computer Cluster Tips

This page provides a range of tips and links for using Underworld on supported computer clusters.

Using Underworld on Computer Clusters

There are broadly three different ways to use Underworld on computer clusters. If you wish to run Underworld models on an Australian supercomputer, it's recommended to use the latest pre-intalled Underworld release module (options 1 or 2), which is installed by the development team and cluster administrators.

  1. Access pre-installed Underworld release modules using a grid submission client. This option does not require a cluster account, although some usage charges may apply.

  2. Access pre-installed Underworld release modules via a traditional shell login. See Underworld Release Modules for further information. This will require a cluster account (see Getting a Computer Cluster Account section below).

  3. Install your own copy of Underworld via a traditional shell login. See Building Underworld on Clusters section below. This will also require a cluster account (see Getting a Computer Cluster Account section below).

Getting a Computer Cluster Account

Contact the site directly to set up an account by following the instructions on their website:

Using Underworld Release Modules on Clusters

Users can utilise pre-installed Underworld release modules, without the need to compile their own copy. See Underworld Release Modules for more details.

Compiling Your Own Copy of Underworld on a Computer Cluster (as at 1.4)

See the Compiling Underworld on Supported Computer Clusters for a set of installation instructions as at the Underworld-1.4 release.

Submitting Compute Jobs to a Queueing System

In your pbs file, include the launch command appropriate to the cluster and version of Underworld you want to use as described below, noting:

  • replace mpiexec with mpirun if using a copy of Underworld which was compiled using Mpich 1 (see Underworld Module Environment for how to check which mpi is being loaded with an Underworld module)
  • "-np X" can be used to specify the number of processors you want to run your model across, replacing X with the number of processors required
  • if you have already loaded an underworld module in the pbs script, you can call the "Underworld" executable as per the example below (the relative or full path to the executable does not need to be specified).
  • if you have compiled your own copy of Underworld, replace "Underworld" with the pathname of the Underworld executable in your own build
  • "InputFile.xml" represents the xml file that you want to run (see Underworld User Manual)
  • if your InputFile and / or a softlink to an executable in your own build of Underworld is located in a different directory to your pbs file, you will need to provide the relative path(s) to these files from the location of your pbs file.
mpiexec -np X Underworld InputFile.xml

See each site's guidelines for more information:

Technical Support

For technical support with Underworld, see Technical Support.

For account queries with a particular clusters, email the site directly:

If you would like to review the capabilites of Underworld or see some examples of the XML format it uses while setting up your first jobs, check out the Underworld User Manual. Sets of template InputFiles can be downloaded from the downloads page for each Underworld release.

-- Wendy Mason - 18 January 2010
-- Julian Giordani- 18 January 2010
-- Kathleen Humble
-- Patrick Sunter
-- John Spencer