...

When it comes to computational analysis or large data transfers, you will rely on compute nodes (and/or Globus /wiki/spaces/RC/pages/349143060 file transfer).

Ways to

...

Get an Interactive Shell

[wdizon@login1wdizon@ood.asre.rc.asu.edu ~]$

This terminal indicates the user is in their HOME directory (~) on login01. From here, we can use the convenience script interactive or the direct command salloc to request and submit requests to allocate compute cores and GPUs.

salloc

$ salloc -c 15 -N 2 -n 2
salloc: Granted job allocation 910
salloc: Nodes ac[001-002] are ready for job
bash-4.4$

salloc is the preferred way to get resources. This is because it encourages all users to be specific about their needs and the syntax is identical to all supercomputers using slurm as the resource scheduler. Both are equally as capable of getting any and all available resources.

interactive

$ interactive
salloc: Granted job allocation 909
salloc: Nodes ac001 are ready for job
[wdizon@ac001 ~]$

With no other parameters specified, interactive will request a single core on a compute node, for one (1) hour.

This can be modified with normal salloc commands, like -t 120 (120 minutes) or -G 1 (allocate a GPU). More details are found in the above link about requesting and allocating above.

salloc

$ salloc -c 15 -N 2 -n 2

...

X11 Forwarding

There is no special switch or option needed for X11 forwarding to work for jobs: it is always enabled.

If you need to run an X11-based program, simply launch an interactive session from either of the aforementioned routes and run the program from within it.

Note, X11 forwarding to your desktop requires you are logged through an SSH client that supports X11 forwarding, such as MobaXTerm on Windows or XQuartz on Mac. Make sure you connect to the supercomputer with the -X option

ssh -X asurite@ood.asre.rc.asu.edu

X11 Forwarding is not always a smooth experience due to X11 being an inefficient protocol; if you experience slowness, you might consider also to use the web portal.