Compute Partitions

There are several different Slurm partitions for different purposes and having different characteristics. The partitions are collected into the following groups based on what they provide:

• CPU Partitions
• GPU Partitions

Which partition to choose?

If you do not request a partition, your job will be placed in the default partition, which is the NHR standard96 partition for all users. This means non-NHR users will have their jobs pending indefinitely with the reason PartitionConfig. Which partitions you have access to is listed on Types of User Accounts. For example new SCC users should specify --partition scc-cpu instead for their non-GPU batch jobs.

Many partitions have variants with a suffix like :SUFFIX appended to the (e.g. medium40:shared). The NAME:test partitions are, as the name suggests, intended for shorter and smaller test runs. These have a higher priority and a few dedicated nodes, but are limited in time and number of nodes.

For NHR users there are NAME:shared partitions, where it is possible to use less than a full node, which is suitable for pre- and postprocessing. All non-NHR partitions are usually already shared and do not enforce a reservation of entire nodes. A job running on a shared node is only accounted for its core/GPU/memory fraction it uses (the maximum of each resource is taken). All non-shared nodes are exclusive to one job, which implies that full compute cost per node are paid, even if fewer resources were requested.

The available home/local-ssd/work/perm storages are discussed in Storage Systems.

An overview of all partitions and node statuses is provided by running

sinfo -r

To see detailed information about a node’s type, run

scontrol show node <nodename>