Difference between revisions of "NCI Guidelines Compute Allocations"
(Imported from Wikispaces) |
|||
| Line 6: | Line 6: | ||
At the Centre, we try to avoid the Centre's projects to run The CMS team can then ask NCI to shift SUs between projects. | At the Centre, we try to avoid the Centre's projects to run The CMS team can then ask NCI to shift SUs between projects. | ||
| − | [[ | + | [[Category:NCI Guidelines]] |
Latest revision as of 01:09, 21 March 2019
Jobs running on the login nodes or on the VDI do not use any compute time allocation. Jobs running on the login nodes have low limits on time and resources they can use. The VDI allows you to have multiprocessor jobs ongoing for up to 7 days without using any of the compute time allocation.
The compute time allocation is measured in SU. You only use compute time when running batch jobs. The amount of SU your job uses depends on the queue and the number of processors. The cost ranges from 1 SU per CPU-hour (walltime) to 3.75 SU per CPU-hour (walltime) depending on the queue the job is submitted. Details can be found on the | NCI user guide.
Once the compute allocation is totally used, the project can use bonus time. In that case, all jobs are automatically moved to the normal queue and get a priority disadvantage. At the Centre, we try to avoid the Centre's projects to run The CMS team can then ask NCI to shift SUs between projects.
