Getting Started on the Lyceum Linux Teaching Cluster

Access to the Cluster

Users with an account can connect to the head node with ssh using the Internet address lyceum.soton.ac.uk. Access via telnet & rlogin has been discontinued for security reasons. See our ssh page for details of how to transfer files to and from your local computer and how to display a GUI from the cluster back to your local machine. Access to the compute nodes on the cluster is controlled by a batch job sytem.

File Store and Disk Space

As the volume of data generated in project work is likely to be much greater than is available on a student's Windows filestore, the teaching cluster has it's own filestore (currently 2TB).  

A users home directory on this system will be created when they first login. The directory will be initially be empty - it's not related to other iSolutions file store (we can give users much more space this way). So if you need to transfer any files from file store elsewhere see the section on "Transferring files to/from a remote host with sFTP" at http://www.southampton.ac.uk/isolutions/computing/hpc/access/ssh.html

Individual users will initially be allocated 20GB of backed-up space for the duration of their projects. We could also consider creating directories for group of users involved in team projects.

Temporary filestore

A total of  400GB of additional temporary storage is available on the head node in the directory /scratch and approximately 100GB on the compute nodes in the directory /tmp. Users can create directories and files in these temporary areas, but this should not be relied on for important data. These areas are not backed up. Files older than two weeks will automatically be removed from the head node /scratch area and may be deleted before this if space becomes a problem - please remember that this space needs to be shared with all other users. Files on the compute /tmp areas will be removed at the end of a user's job (so if still needed, they would need to be copied back to your home filestore before your job finishes).

Editing and viewing files, working with directories

If you do not have any previous experience of using command line editors in Linux you will probably want to use the emacs or gedit text editors to modify files. The command emacs my_file displays a new window in which you can view and modify the contents of the text file my_file in an intuitive manner, without having to know anything about the many more powerful capabilities of advanced emacs use.  (Note that you must be able to display new windows back to your local system - see our page on displaying a GUI. If you want to use a simple command-line editor in your original login window then nano is a good choice.)

To list which files and subdirectories you have in a directory use the ls command. (Directories are the equivalent of windows folders.)

To view or search the contents of the text file some_file use the command less some_file. To learn how to move around in the file or search with less type the letter "h" for help while in less.

Use the mkdir command to create a new directory, eg. mkdir new_dir to create a directory called new_dir. To change the directory use the cd command, eg. cd new_dir. Two dots are used to represent the directory above, so cd .. will change to the directory above. The pwd command will show you where you are in the directory structure.

Further information can be found on our introduction to Unix page.

Running Jobs

The head node is designed primarily for shortish interactive jobs of upto an hour or so and for basic testing of how to run applications in batch mode. Jobs which require runtimes of much longer than an hour should be run on the compute nodes in batch mode. To prevent the head node being overwhelmed by long-runing jobs any processes that use much more than an hour of CPU time will be killed automatically. There are 16 computational nodes which can be accessed by submitting a job to the  the PBS/Torque batch job system. Most jobs will be run in batch mode, though interactive jobs are possible. If all the compute nodes are currently busy the job will be put into a queue and the order in which jobs are then run is determined by the Maui job scheduler. More information on how to submit jobs is given in the Job submission overview.

Compiling

The use of the Portland Group compilers is recommended for Fortran codes. For serial codes these are accessed as pgf95 & pgf77 for f90 and f90.The Portland C compilers are pgcc and pgCC, for  C and C++ respectively. The gnu compilers gcc and gfortran are also available and perhaps more portable.  Note that by default you will get the version of the gnu compilers that is packaged with the operating system, which is 4.1.22. The more recent 4.3.3 version of the gnu compilers is also available via the module command which is described in the next section.

Software and Application Packages

See the software page for a list of the main applications, compilers and libraries installed on the cluster. This page will also have links to more information on using a package. The user environment for most of these packages can be conveniently set up with the module command, eg. before you first use matlab in a session:

module load matlab 

Thereafter you can run matlab as many times as you like. If you get get a message that an application can't be found (eg. matlab: Command not found) it generally means that the module is not loaded (or perhaps you've mis-typed the name of the application!).

The  modules command can be used to check which applications are available through modules, eg.

module avail

You can check whether a specific package is available, eg. for gcc:

module avail gcc

Modules can also be used to get a brief summary of how to use the application, eg.

module help cfx

Note that some modules will be loaded for you when you first login (eg. compilers) if you are likely to be using an application everytime then you can ask for the appropriate module to be loaded for you everytime you login with the module initadd command, eg.

module initadd fluent

If you want to check which modules you have loaded at any time use:

module list

Parallel Environments

Many applications such as Fluent or CFX have their own provision for running in parallel over multiple processor-cores (if sufficient licences are available), details are given in the software pages for each application. It is also possible to run Fortran or C codes in parallel if you already have a suitable code using either MPI or shared-memory (but producing a parallel code from scratch is normally a lengthy and intensive process). See the MPI page (in preparation) for more information on using MPI on the cluster.

User Support

System problems and general queries should be directed to ServiceLine.