Partitions
Below is a list of partitions, along with their time limits and purposes. The number of nodes listed in each table is subject to change at any time based on demand and may not be fully accurate.
BOSE
| Partition Name | # Nodes | Time Limit | Purpose |
|---|---|---|---|
| week (default) | 56 | 7 days | General use partition. It should be used when your job will take less than a week, or if you can restart your job to continue running it from a checkpoint. This partition has the most nodes and is highly recommended for most jobs. |
| month | 15 | 30 days | Special partiton for longer-length jobs, but nodes are shared with the week partition. |
| GPU | 4 | 7 days | Partition that uses exclusively nodes that contain GPUs, only to be used when GPUs are required for your job. Note: You must specify –gpus=# to indicate how many GPUs you want to use. |
| highmemory | 3 | 7 days | Partition for jobs that require large amounts of memory to compute. Our two nodes support up to 990GB of memory. |
| medea | 52 | 7 days | Specialty partition for jobs that use the MedeA software. |
| pre | ALL | 24 hours | This is a low priority special partition only to be used for testing and debugging your software. Jobs that use this partition will run on any available node with space, but will be requeued if a job comes in using another partition. |
BGSC
| Partition Name | Time Limit | Purpose |
|---|---|---|
| week (default) | 7 days | General use partiton. It should be used when your job will take less than a week. This partition has the most nodes. |
| batch | 30 days | Special partiton for longer-length jobs. |
| GPU | 7 days | Partition that uses exclusively nodes that contain GPUs, only to be used when GPUs are required for your job |
| extended | 104 days | Special use partiton for extremely long jobs. If more than 104 days are needed, please contact the BGSC admins |
| scavenge | 5 days | Note: it is STRONGLY recommended to not use scavenge for important jobs. Scavenge is a special partition that has access to every node, and will use any available resources. However, if any job running on a different partiton requires resources that a job on the scavenge partition is currently using, the job on scavenge will be stopped. |
Using A Partition
By default, both clusters will assume you want to use the 'week' partition if you do not specify one when you submit your job. If you would like to use another one, such as GPU or highmemory, you'll need to make sure that either your Slurm script references your requested partition, or you specify it when running 'sbatch'.
Slurm Script
You can also specify your partition, or any other Slurm option, directly when you submit your job. This will override what you put in your script file.
Slurm Interactive Mode
You can also use "--partition=partitionhere" when running interactive mode if you wanted to type commands manually on one of our compute nodes.
Want to use GPUs?
If you want to use one of our graphics cards in your job, you'll have to set both the partition to "GPU" and specify how many GPU cards you want to use.
Slurm Script
#SBATCH --partition=GPU # Set partition to use GPU nodes
#SBATCH --gpus=1 # Request one GPU, for a max of three.
Interactive Mode