You can access Roar and Roar Collab either by purchasing an allocation or by using the “open queue” for either system. The open queue is a free, shared resource for all researchers, whereas paid allocations have varying prices and include access to varying resources (view rates).
Want to try out an allocation before you buy? Request a free 30-day trial allocation.
A paid allocation lets you access the resources you need for your research within a time frame that you choose. With a fully executed Service Level Agreement (or SLA), any jobs that you submit within your allocation limit are guaranteed to start within one hour of when you submit them.
Other benefits include:
- No wall time limit
- Burst capability (use up to 4x your allocated compute for a limited period)
- 5 TB of Group (active) storage with minimum purchase
With a paid allocation, you can choose to purchase access to several different core types:
- Basic Core
- Standard Core
- High-memory Core
- GPUs (two types—NVidia A100s and P100s)
You can also purchase additional storage or archive storage for your research group.
To request a paid allocation, contact us.
Learn more about services and rates (requires Penn State login).
The Open Queue
Researchers who do not have a paid allocation can access Roar and Roar Collab through their respective “open queues.” All you need is an Roar or Roar Collab account. If you don’t have one, request an account here.
Jobs submitted to the open queue run using idle computing resources. That means the jobs you submit to this queue do not count against your paid allocation.
- All users have equal priority for open queue access
- Open queue jobs have no guaranteed start time
- Jobs and interactive sessions will start and run only when sufficient idle cores are available
- Jobs may be cancelled if cores are needed for jobs on other queues (when submitting you may specify a preference to restart pre-empted jobs)
For each account, the open queue has these maximum limits:
- 100 jobs pending
- 100 cores executing jobs at any given time
- 800G total memory across all jobs
- 48-hour job wall-times
- 24-hour interactive session durations
If you have any questions about using the open queue or submitting jobs, please contact email@example.com.