GPU Remote Workstation Booking System

GPU Remote Workstation Service (Experimental)

 

We provide access to remote GPU workstations that you can use via your web browser.

The machines are equipped with NVIDIA GeForce RTX 4070 Ti SUPER or 4090 GPUs and are available to reserve time on.

When you reserve time on one of these machines, a remote access session will be created for you, and you will have scratch space on the machine’s disk in /scratch0/$USER/

This scratch space and any other data on the disk will be removed when your sessions ends, so you must ensure that you save any work you want to keep in your home folder or project store.

Making a Booking

    1. Login to the booking system at https://mydesk.cs.ucl.ac.uk using your Computer Science account username and password. Email login is not supported.
    2. Click on the Reservations menu, then select Schedule to go to the view of all schedules.
    3. Click on the dropdown menu at the top of the page and select GPU Workstations.
    4. You can now click on a time slot for a machine to reserve it.
      • On the New Reservation page, enter a title then click the Create button.
    5. Once you have made a reservation, a remote access session will initiated for you on the machine at that specified time.

Accessing Your Booked Machine

    1. At the time of your booking, in a web browser, go to the following URL modified for your reserved machine’s host name:
      https://<host>.cs.ucl.ac.uk:8443/guacamole

      • For example: https://blurr.cs.ucl.ac.uk:8443/guacamole
    2. Login to the Apache Guacamole remote session using your Computer Science username and password.
    3. You will see a desktop environment running Rocky Linux and your department home folder will be available.

Note: If you log out this will end your graphical session early!

If you need to, you can close your browser tab and open it again later.

Session End

Prior to your session ending, you will receive a warning to save any work you have on the machine to another location.

At the end of your reservation, your Apache Guacamole remote session will automatically terminate, and any data will be cleaned off the machine for the next user.