Job Partitions and Sizing
-------------------------

.. _`partition`:

Partitions
^^^^^^^^^^
Jobs are submitted to partitions (queues), which control which resources can be used by the job and other restrictions such as wall time. Some resources can be used by jobs from more than one partition.

The partitions available are:

+---------------+-------------------+-----------------------------------+
| Partition     | Max wall time     | Comment                           |
+===============+===================+===================================+
| test          | 30 min            | Use for debugging your scripts    |
+---------------+-------------------+-----------------------------------+
| short         | 6 hrs             |                                   |
+---------------+-------------------+-----------------------------------+
| medium        | 24 hrs            |                                   |
+---------------+-------------------+-----------------------------------+
| long          | 2 weeks           | (research accounts only)          |
+---------------+-------------------+-----------------------------------+
| gpu           | 1 day             | GPUs (restricted)                 |
+---------------+-------------------+-----------------------------------+


Job Sizing
^^^^^^^^^^

The cluster machines have the following specification:

- ``fury``: 48 cores, 128 GB RAM
- ``hornet``, ``dragon``, ``gauntlet``: 56 cores, 256 GB RAM
- ``mohawk``: 64 cores, 256 GB RAM

The maximum number of cores which can be requested is therefore 64 cores.
The minimum number is 1.

It recommended to choose the number of CPU cores based on known scaling of
the problem / code, but to be a multiple of 8 where possible, to improve 
the allocation of resources.