 |
queue
name |
min
#nodes |
max
#nodes |
max
wallclock |
priority |
notes |
| debug |
|
2 |
15 min |
high |
"always available"; use for interactive jobs |
| ib |
2 |
100% |
1 week |
high |
Infiniband connected nodes |
| wide-debug |
5 |
100% |
15 min |
high |
|
| narrow-med |
|
20% |
8 hr |
med |
|
| wide-short |
5 |
100% |
2 hr |
med |
|
| narrow-long |
|
20% |
3 day |
low |
|
| narrow-extended |
|
25% |
2 weeks |
low |
ICP accounts only (paid allocations) |
| med-extended |
5 |
50% |
1 week |
low |
ICP accounts only (paid allocations) |
| wide-med |
5 |
100% |
8 hr |
low |
|
| serial |
|
100% |
unlimited |
very low |
preemptible. No SUs charged. |
Goals, general principles, and observations considered by the HPCC-AAC:
- Allocation approvals include total compute-hours and allowable queues.
- Preference is given to big, parallel jobs (and development/testing to prepare these.)
- All serial jobs must be submitted via qsub -q serial, have low priority, are preempted.
- Special requests are considered in exceptional cases.
- Installed software base and versioning are controlled by the HPCC-AAC.
- Jobs with very large data and/or setup times may need additional considerations, tools.
- Some researchers will need assurance of security for their data on the cluster.
- Grants requiring a security checklist should contact the HPCC-AAC.
HPCC-AAC = High Performance Computing Cluster
Allocations and Advisory Committee
|
