Slurm Partition Overview
The Discovery Cluster utilizes several Slurm partitions to allocate computing resources for different workloads. Each partition may have multiple entries indicating current availability, time limits, number of nodes, states, and node lists.
Below is a detailed breakdown of the partitions from the current cluster status (sinfo output):
Partition | Time Limit | Description | Nodes |
---|---|---|---|
standard | 30-days | Default partition for general use | q01-03,05,09,s01-29,43-44,t03-12 |
gpuq | 10-days | Partition for GPU-related jobs. | a01-a02 |
ood | 3-days | Partiton for Open OnDemand | a02, s41 |
v100 | 10-days | Partitions with V100 GPUs | p01. |
a5000 | 3-days | Preemptable partition for all users | amp01-03 |
a5000_w | 3-days | Preemptable partition for all users. | amp05-06 |
a5500 | 3-days | Preemptable partition for all users. | centurion01-05 |
a5500_w | 7-days | Preemptable partition for all users. | centurion06-09 |
a100 | 3-days | Paid tier for access to A100 GPUs. | a03-a05 |
preemptable | 30-days | Partition for general use | q10,r01-02,04-21,s01-19,22-27,30-40,t01-12 |
v100_preemptable | 3-days | Private partition for vaickus group | gv01, p02-04 |
hautier_high | 10-days | Private partition for the Hautier group | r01-02,04-21 |
hautier_low | 10-days | Private partition for the Hautier group | s30-40,42 |
preempt_robust | 7-days | Private partition for Robustelli group | amp01-04,centurion01-05,turing01-02 |
preempt_wenlin | 3-days | Private partition for Wenlin group. | amp05-06,centurion06-09 |
l40s_nova | 3-days | Public partition for free users limit 2 GPUs | adanova01 |
l40s_indrani | 7-days | Private partition for bhattacharya_lab. | adanova01 |
preempt_indrani | 7-days | Private partition for bhattacharya_lab. | adanova01 |
preempt1 | 30-days | Private partition for dac | q10 |
preempt_hautier | 10-days | Private partition for the Hautier group | r01-02,04-21,s30-40 |