This information is not valid for users of the Red Española de Supercomputación (more information).
In order to improve the scheduling and reduce job’s waiting time, a structure of partitions is defined.
The following table summarises all the relevant information of every general available partition.
Partition | SLURM directive | #Nodes | Nodes | Cores / node | Threads / core | Mem (MB) / core | Limits | Information |
---|---|---|---|---|---|---|---|---|
std | #SBATCH -p std | 57 | pirineus[7-62] | 48 | 1 | 3900 | - | The default partition, with standard nodes. Most jobs will use it. |
std-fat | #SBATCH -p std-fat | 6 | pirineus[1-6] | 48 | 1 | 7900 | - | Partition that includes standard nodes with more memory per core. Designed for jobs that require between 3900 MB and 7900 MB of memory per core. |
mem | #SBATCH -p mem | 2 | canigo[1-2] | 192 | 1 | 24064 | - | This partition includes the shared memory machine named canigo. |
gpu | #SBATCH -p gpu | 4 | pirineusgpu[1-4] | 48 + 2 GPGPU | 1 | 3900 | - | Partition with nodes that are equipped with P100 Nvidia GPGPU. |
knl | #SBATCH -p knl | 4 | pirineusknl[1-4] | 68 | 4 | 5600 | - | Partition with nodes equipped with 2nd Generation Intel® Xeon Phi™ Processors. Designed for jobs that require the use of Knight's Landing architecture. |
express | #SBATCH -p express | 1 | pirineus63 | 48 | 1 | 3900 | 1 hour and 4 cores per job | This partition is intended for test jobs, either interactive or batch. |
Related articles