You may use up to 350 CPU cores or 1TB of RAM on the cluster compute nodes. Beyond these limits your work will queue. Once your running jobs complete your queued jobs will start.
You are welcome to run smaller jobs directly on the Acropolis head node. The head node is a great place to develop your application. Feel free to test your work using a few cores. Once you are ready to scale up your work you may submit it to the cluster nodes.
The head node is not an appropriate place for large multi-core or large parallel applications. Programs using more than a few cores will be throttled. Parallel applications may be paused if they present a risk to system stability or appear to be running out of control.
Some graphical programs can be launched on the cluster nodes from the start menu in a VNC session. Programs such as Matlab, Rstudio Desktop, Xstata will ask you to reserve CPU/RAM before launching on the cluster nodes.
Please be sure to save your work and close the program once you have completed your tasks. Idle programs on the cluster will be set preempt-able. If the cluster is busy preempt-able programs may be terminated in order to allow newly submitted jobs to run. You will be notified if your program becomes preempt-able. If you return to work on a program which has been set preempt-able it will not be terminated.
When the head node is under moderate heavy load you will be prompted with a choice to launch programs on the compute nodes. If the head node is under extreme load programs will automatically launch on cluster nodes.