Job Limits
These are open for review if you find any of them unreasonable or inefficient.
Info
This page is about hard limits applied to jobs and users. For more information about per project usage limitations, see our page on Fair Share.
Per Job¶

- 10 nodes
- 21 node-days
- 21 days
Per User¶
- 2688 CPU cores occupied (so 16 full Genoa nodes, or 21 Milan nodes),
- 3528 core-days booked by running jobs (so 3 weeks of one full node).
- 6 TB of memory occupied (4 full 1.5 TB nodes)
- 30 TB-days booked by running jobs (so 3 weeks of one full 1.5 TB node).
- 6 GPUs occupied, 14 GPU-days booked by running jobs (so 2 GPUs for 1 week).
- No user can have more than 1,000 jobs in the queue at a time.