In order to better manage the computing resources available on the GPC, the following individual usage policies are in effect:
- Users are restricted to a limit of 25 slots at a time. This effectively means that parallelized jobs should not request more than 25 slots at a time, or they will not be able to move out of the queue.
- A "fair share" weighting system is in place on the job scheduler. Whenever there are more jobs requested than the system can accommodate, priority will be given to jobs from users who have recently used fewer resources. This meter will refresh on a weekly basis.
- When the scratch file system is more than 70% full, users whose jobs use more than 50 GB of scratch space will be notified with a request to ask them to stop their job.
- Users are limited to 1 qlogin session at a time, which can run for no more than 24 hours. Users who need to use GPC resources for longer than 24 hours should do so by submitting a batch job to the scheduler (see https://computing.sas.upenn.edu/gpc/job for more information).