Skip to content

Chalmers BSc and MSc courses on Vera

If you are teaching a course, MSc, or BSc thesis at Chalmers you may apply for CPU, GPU, and storage resources on the Vera cluster.

You will get:

  • Access to Vera slurm cluster for interactive and batch jobs for teachers and all students.
    • Automatic limits on how much individual students can use
  • Group storage on Cephyr accessible by all project members for sharing software and data.
  • Technical support for teachers (not students) for making jobscripts, software installs, etc.
    • Teachers are resonsible for helping their students
    • Software module tree and containers, with custom containers
  • Teachers and students log in with their CID
    • SSH and remote desktop to login nodes
    • OpenOndemand portal for launching jupyter notebooks, remote desktops, and more on compute nodes.
  • Short lives accounts: student home directories will be cleared out 4 weeks after course ends.
  • Short queue reservations for scheduled lab session upon request.

Compute time on Vera is funded by GRU.

Members

SUPR is not used, instead the teacher instead sends a CSV file with all students to support@c3se.chalmers.se once the project is approved. You can export such a CSV file from Canvas (Course page → Grades → Export entire gradebook view), and you must submit send this to the support at least 5 working days in advance.

  1. Students need to pass Introduction to computer clusters This is checked by the teacher of the course.
  2. Students should ask their teacher for support.

Applying for access

In the near future we hope to have a website up where you can apply, but in the meantime you apply via email to Mikael Enelund who handles approvals:


For BSc, MSc, and projects that don't have a unique course codes please state so.

Note that allocations on CPU and GPU are separate, if you don't apply for GPU hours your project won't be able to submit GPU jobs at all and vice-versa. Cost for a GPU-hour varies slightly by model.

Examples:

  • 50 students doing light AI work, ~4 hours per student / month = 200 GPU hours / month
  • 10 students running CFD simulations across single node jobs 64 cores for a total of 10 hours / month = 7 x 1000 core-hours/month
  • 50 students and each student shouldn't be able to exceed 2 times more than their fair share means per student limit = 4%