Cluster manager on ACISS has a number of queues available to users.
- Normal queues (1 day time limit):
- generic(128 nodes, 12 cores & 72 GB RAM per node)
- fatnodes (16 nodes, 32 cores & 256-512 GB RAM per node)
- gpu (52 nodes, 12 cores & 72 GB RAM & 3 nvidia M2070 GPUs per node)
- Additional queues:
- short: generic node for 4 hours
- student: generic nodes for students only (1 day)
- longgen, longgpu, longfat: generic, gpu and fat nodes w/ 4 day default limit and 2 week maximum
To access additional scratch on fat nodes, append :scratch to your resource requests, e.g.:
qsub -q fatnodes -l nodes=1:scratch
If you need a fatnode with extra RAM, include the “xmem” resource option:
qsub -q fatnodes -l nodes=1:ppn=32:xmem
The queues are subject to change depending on needs. Let us know what we can do to better meet your needs. If you have any suggestions, or questions or need any assistance, please send us an email to firstname.lastname@example.org
To open interactive sessions, use qsub -I. To get an interactive session with graphics forwarding, qsub -IX. Successful graphics forwarding requires that you also logged into the head node with graphics forwarding also, e.g. ssh -Y email@example.com.
To learn more on submission queues, check out the HowTo.