IMPORTANT 1 – RUNNING CODE ON THE HEAD NODE
We have noticed an increasing number of people running software or installing software\libraries directly on the head node. Please do not do this. The head node is a small, virtual machine that is only intended to run the HPC scheduler. Please use an interactive job to install software or libraries. If you are uncertain how to do this please contact us and we will gladly assist.
We have noticed an increasing number of people running software or installing software\libraries directly on the head node. Please do not do this. The head node is a small, virtual machine that is only intended to run the HPC scheduler. Please use an interactive job to install software or libraries. If you are uncertain how to do this please contact us and we will gladly assist.
IMPORTANT 2 – GPU USAGE
The GPU cards are currently in very high demand. Please ensure that your GPU card usage is in-line with your gres reservation. The University cannot afford to have these high value assets blocked by over-subscription. Repeat offenders will have their GPU card limits set to single card usage or be removed from the GPU partition entirely.
vscode server
For those who make use of it, do not run vscode server directly against the head node. If you need to run vscode server you have two options:
1 – point your vscode server instance at the worker node where your sbatch script or interactive job is running.
2 – login to On Demand and launch a web based vscode server job. You can even disconnect your browser and reattach to the job at a later time.
On Demand
We have implemented On Demand, a graphical, web based service that will allow researchers to run various packages on the cluster via a browser. In addition On Demand allows command line shell access, file transfers and basic file editing. The system is stateful, so once a job is launched you can shut down your browser and reconnect to the same job at a later time.
Several packages have already been ported to On Demand such as Ansys, Matlab, R Studio Server, Mathematica and Jupyter Notebooks.
New servers
We have recently added another sixteen 48 core ada:100 series nodes and plan to add another four soon. We have also added an additional GPU server with four L40S cards.
HPC Training courses
We will be running an Introductory HPC training course on Thu 25th of September. If you are interested in attending please let us know. The course will be a morning session at the ICTS offices.
We will be running an Introductory HPC training course on Thu 25th of September. If you are interested in attending please let us know. The course will be a morning session at the ICTS offices.