OpenStack
Dashboard is available at os.ai.ii.pw.edu.pl if you are connected to the VPN. Read the User Manual to get started.
Explanation
This table shows the current availability of graphic cards. Each line shows the number of instances that can be created with a specific flavor. Flavors are named in the form of GPU-NODE.
How to use this table? Look at the numbers in the second column. This is the availability NOW. Select a graphic card – node pair you want from these that have current availability greater than zero.
After you create a new instance you will see the availability of a selected flavor drop by one. It may take a couple of seconds until the numbers change.
Documentation
Rules
- One GPU-enabled instance per person.
- Use volumes for persistent storage. Instances must be ephemeral.
- Delete your instance when you will not use it for more than a day.
- Delete unneeded volumes.
Key differences
This new OpenStack differs from the previous one at legat. The key differences are:
- 18 graphic cards. (previously: 9)
- 1120 GB RAM. (previously: 608 GB)
- New node: H81. (RTX A5000, AMD EPYC 7313)
- New, reliable, real-time free instance table. (see above)
- The recommended way to free resources is to delete, not shelve instances. *
- The Configuration Drive checkbox is selected by default.
- The Security Groups are configured for SSH access by default.
- Selecting an SSH key is now required when creating an instance. **
- Automatic enforcement of the resource freeing policy
* This way you can easily create a new instance of an arbitrary flavor. No more issues with unshelving if there are no more available instances of selected flavor. Deleting an instance does not delete data, because data is persisted on volumes.
** This prevents creating an instance that cannot be accessed via SSH.