Types of Nodes & Policies#

Login Nodes#

The login nodes are your gateway to the cluster from which you view/edit/upload files and dispatch jobs to the compute nodes. These nodes will be accessible over SSH however running any intensive programs is forbidden.

Compute Nodes#

The compute nodes are where your jobs will run. The compute nodes mount all shared filesystems making software and files available for your jobs irrespective of the node/s in which they run. These nodes are not accessible over SSH and direct access via methods other than the scheduler is forbidden.

GPU Nodes#

GPU nodes are principally the same as compute nodes but with the addition of special compute optimised GPUs typically used for accelerating modelling of engineering applications of AI / machine learning tasks. Jobs requiring the use of GPUs must specify this requirement as part of their resource request.

Large Memory Nodes (Currently Unavailable)#

In the future we will add large memory - nodes that are identical to normal compute nodes but have additional memory available to ensure they are capable of running more memory intensive SMP / OpenMP jobs. Jobs requiring sufficiently large amounts of RAM will automatically be dispatched to these nodes.

Reserved / Specialist Nodes (Contact CS Department)#

In the future, we see interdisciplinary research requiring these reserved nodes. Suppose your department / research project has the funding to purchase compute infrastructure, we propose you add your nodes to the university cluster as reserved nodes.

Reserved / specialist nodes are nodes not available in the public ‘free’ queues and have either standard or non-standard specifications. They are reserved for the exclusive use of the purchasing users/department/research groups. Some of these compute nodes will be purpose built to accelerate specific tasks and may have different job scheduler queue configurations.

Jobs requiring or desiring the use of specialist nodes must specify this requirement as part of their job resource request.