Partition List¶
Core Limits¶
There is currently a 300 CPU core, 64 GPU limit to be shared by the users of each lab.
When you try to go over this limit, you will be denied for MaxCpuPerAccount.
General Use Partitions¶
These are the only partitions that are open to use by all users of Unity.
| Name | Relative Wait Time | Default Job Time | Time Limit | Max CPU's Per Node |
|---|---|---|---|---|
| cpu | medium | 1 hour | 1 day | 40 |
| cpu-long | long | 2 days | 14 days | 40 |
| cpu-preempt | short | 1 hour (see below) | 14 days (see below) | 256 |
| gpu | medium | 1 hour | 1 day | 72 |
| gpu-long | long | 2 days | 14 days | 72 |
| gpu-preempt | short | 1 hour (see below) | 14 days (see below) | 64 |
| power9 | short | 1 hour | 30 days | 128 |
| power9-gpu | medium | 1 hour | 30 days | 128 |
| power9-gpu-preempt | short | 1 hour | 14 days | 128 |
| arm-preempt | short | 1 hour | 14 days | 80 |
Preempt¶
Nodes in preempt partitions are buy-in nodes which would normally be inaccessible to general users. Buyers in reserve the right to kill (preempt) any job running on their nodes to make room for their own jobs. When you schedule a job on a preempt partition, your job can be killed after two hours.
If your job is preempted, it will be re-queued and start again on some other hardware. If you don't want your job re-queued (but still killed), you can use the --no-requeue Slurm argument. To avoid losing progress, it's a good idea to use software that supports checkpointing. Checkpointing software periodically saves its state to a file. In the event of an interruption, it can read from this file and resume work where it left off.
Gypsum Cluster Partitions¶
Gypsum users, depending on the type, have access to these partitions.
| Name | Time Limit | Max GPUs - Other | Max GPUs - MS | Max GPUs - PhD | Comments |
|---|---|---|---|---|---|
| gypsum-m40 | 7 days | 1 | 6 | 12 | M40 GPU partition |
| gypsum-titanx | 7 days | 4 | 20 | 40 | TITAN X GPU partition |
| gypsum-1080ti | 7 days | 4 | 20 | 40 | 1080 Ti GPU partition |
| gypsum-2080ti | 7 days | 2 | 10 | 20 | 2080 Ti GPU partition |
| gypsum-rtx8000 | 7 days | 0 | 3 | 6 | RTX 8000 GPU partition |
IALS Cluster Partitions¶
IALS members have access to these partitions.
| Name | Time Limit | Comments |
|---|---|---|
| ials-gpu | 14 days | GPU partition for IALS |
Other Priority Partitions¶
These partitions represent the specific labs who "bought in" to Unity by providing their own nodes.
| Name | Time Limit | Comments |
|---|---|---|
| ceewater_cjgleason-cpu | Unlimited | ceewater CPU partition for cjgleason group |
| ceewater_casey-cpu | Unlimited | ceewater CPU partition for casey group |
| ceewater_kandread-cpu | Unlimited | ceewater CPU partition for kandread group |
| astroth-cpu | Unlimited | astroth CPU partition |
| zhoulin-cpu | Unlimited | zhoulin CPU partition |
| toltec-cpu | Unlimited | toltec CPU partition |
| gaoseismolab-cpu | Unlimited | gaoseismolab CPU partition |
| uri-cpu | 30 days | URI CPU partition |
| uri-gpu | 30 days | URI GPU partition |
| ece-gpu | 5 days | ece partition for ECE courses only |
| umd-cscdr-cpu | 30 days | UMD CSCDR CPU partition |
| umd-cscdr-gpu | 30 days | UMD CSCDR GPU partition |
| umd-cscdr-arm | 30 days | UMD CSCDR ARM partition |
| power9-gpu-osg | 30 days | OSG Resource Contribution |