On the Supek cluster, there are several job queues that are divided depending on the availability of resources and the time of job execution:
Name | Description | Available resources | Limits | Priority | TMPDIR |
---|---|---|---|---|---|
cpu | Standard queue for CPU nodes | ncpus=6656 - Total ncpus=128 - per node mem=256GB - per node | max 7 days | 1 | /lustre/scratch |
gpu | Standard queue for GPU nodes | ncpus=1280 -Total ngpus=80 -Total ncpus=64 - per node ngpus=4 - per node mem=512GB - po čvoru | max 7 days | 1 | /lustre/scratch |
bigmem | Standard queue for nodes with large memory capacity | ncpus=256 -Total mem=8192GB ncpus=128 -per node mem=4096GB - per node | max 7days | 1 | /scratch (1,7TB) |
cpu-test | Queue for testing on CPU nodes | ncpus=6656 -Total ncpus=128 - per node mem=256GB -per node | max 1 hour max 256 jezgara i 500 GB RAM per user | 2 | /lustre/scratch |
gpu-test | Queue for testing on GPU nodes | ncpus=256 ngpus=16 | max 1hour max 4GPUs i 500 GB RAM per user | 2 | /lustre/scratch |
login-cpu | Queue for testing on the CPU login node | ncups=100 | max 1hour | 1 | /scratch (1,6TB) |
login-gpu | Queue for testing on the GPU login node | ngpus=1 | max 1hour | 1 | /scratch (3,4TB) |