Quick start guide
To log in on PDC's Cray XE6 system Lindgren you need to use. The login node is called:
Note that the login nodes on the Cray are quite weak as they have just 6 cores, and do not have any disks or paging space, and only have 16 GB of RAM available. If the available RAM on a node is used up, it is likely that the node will become unresponsive and will have to be rebooted. So, if possible, avoid running any processes that consume large amounts of CPU time or memory on the login nodes. Instead, such tasks should be run on the interactive nodes, or on a different machine (such as Ferlin's interactive nodes).
When you log in to Lindgren, you will arrive in your PDC AFS directory:
Your AFS directory is a good place to keep back-up copies of important files. However, files in your AFS directory are not available to the Lindgren compute nodes. This means that any data that is located in your AFS directory cannot be accessed by jobs running on the Lindgren system.
There is another file system, called Lustre, that is used to access files on the compute nodes. Your Lustre space is available at:
You can compile your programs in your AFS directory, but you must run your programs from your Lustre space, since this is available on all the compute nodes (whereas your AFS space is not). Note that the Lustre file system is not backed up, so it is a good idea to keep a copy of any important files in your AFS directory.
REMEMBER that your PDC AFS home directory:
(and any other volumes that you might have in AFS) are not available on the Cray's compute nodes. Therefore, if there any files you want to use that are currently in your AFS space, you need to copy them to the Lustre file system.
Also, as AFS is only available on the Cray login node, you should not create a symbolic link to your Lustre directory from your home space. Even though the final result would be a directory on Lustre, the compute nodes would not be able to find the link to determine that.
If you need to transfer larger amounts of data, please use the dedicated transfer node cfs-aux-4.pdc.kth.se instead of the login node!
At this point you might like to read a bit about Lindgren's hardware. Knowing how the machine is actually built helps you to to use the computer more efficiently.
To find out how to access software on Lindgren, see the software section.
Compiling a program
In general, the modules that you have loaded will govern the choice of the compiler and libraries that will be used when you compile a program on Lindgren. You need to compile for the compute nodes where your program will be run. It is important to be away that the software environment on the compute nodes and the login node are very different.
Running a program
When your program has been compiled for the Cray XE6's compute nodes, and when the compiled program has been placed on the Lustre file system, it is time to run the program.
When you are developing your code or if you need to quickly test run something, there are compute nodes available for interactive use on the Cray. Please note that you should only run processes of maximum one hour in duration on these interactive nodes!
The batch mode compute nodes are accessed through the queuing system, whereas these interactive nodes are accessed directly from the login node using aprun (which is roughly equivalent to mpirun on the Cray). The resulting job will then be launched on the interactive nodes.
When you have read the documentation about Lindgren's hardware and how to compile and run programs on the system, you might also find it useful to look at some hands-on examples showing how to run simple MPI and OpenMP programs on the Cray.
Best Practice Guide
Additional information about using Cray XE systems is available in the PRACE Best Practice Guide - Cray XE.