The purpose of this page is to provide the new user with the bare minimum of information required to start using CHPC's facilities. Please explore the rest of the Wiki for more advanced information, as well as application-specification instructions and examples.

1. Most high performance computing (HPC) systems like CHPC's cluster use the Gnu/Linux operating system. Windows applications will not run on the system.
2. The computers that make up the HPC cluster do not run faster than a good desktop computer, nor do they usually have particularly impressive amounts of memory. The performance comes from running many compute nodes in parallel. If your application is not parallelised, it will not run faster at CHPC than on a good desktop machine.
3. The system is designed to be used in batch mode. The user prepares a job, submits it to the queue, and the system executes it when the requested amount of resources become available.
4. You can request an interactive session, which will give you access to a compute node in the cluster. Here you can compile programs, do pre- or post-processing, and check your job setup.
5. Access to the system is by means of secure-shell (ssh).
6. Data can be copied to or from the system by means of secure copy (scp) or rsync.
7. Tasks are managed with PBS
8. Although tasks are run in batch mode, the visualisation server makes provision for graphical pre- and post-processing.
9. Software licensing issues still need to be taken care of appropriately. CHPC hosts academic licenses for some software suites, but for others you will need to make the necessary arrangements with the software vendors.

Once you have a user-id and password, you can log in to the system using ssh. If you are logging in from a system running a UNIX-like operating system, you can simply use ssh from the command line: ssh myusername@sun.chpc.ac.za. From Windows, you can log in using Cygwin or PuTTY. Once logged in you can (and should) change your password with the passwd command. Options for logging in are explained in the following guide.

Although personal computers running Linux have user interfaces that are very similar to Windows or Mac, the ssh-login to the HPC cluster runs in command line text mode only, by default using the Bash shell. It is therefore essential to master a very small subset of Linux commands. The internet has many excellent resources explaining how to perform tasks in Linux and Bash. Nevertheless, here is a very short summary of the most useful instructions:

• ls - provide a list of the files in the current directory
• ls -l - provide a more detailed listing of the files in the current directory
• pwd - gives the current directory path (Present Working Directory)
• mkdir - create a new directory
• cd - change to a new directory
• rmdir - remove a directory
• rm - remove a file
• nano - start the nano text editor (recommended for beginners)
• vi - start the vi text editor (warning: steep learning-curve)

Although some basic software is immediately available on the cluster, most of the more useful software packages need to be accessed via the modules system. This includes newer and more capable versions of the gcc and intel compilers, various python versions with numpy, etc, many software libraries and specialized scientific packages.

To list all available packages, use the following command:

module avail

Most of the useful modules are part of the chpc modules group. You can list just these as follows:

module avail chpc

In a similar way, you can refine the search further, eg.

module avail chpc/earth

That will list all the applications and libraries in the “earth” subgroup (reserved for earth-science applications. chpc/compmech contains the computational mechanics packages (CFD, FEA, etc)

To actually load a module and be able to use the software, you should use the module load or module add command (they do the same thing), eg.

module add gcc/6.1.0

module load chpc/python/3.5.2_gcc-6.2.0

To list all the modules that you have loaded, type:

module list

To unload a module, type:

module dell gcc/6.1.0

and finally, to unload all modules, type:

module purge