Skip to content

Remote visualization

Remote data visualization service allowing pre- and post-processing using graphic software

Prerequisite: Install the latest version of the TurboVNC client on your workstation [].

Step 1: Start a graphical session

**From a Austral frontend node: request a graphical visualization session by typing the command: startvisu.

Help is available by adding --help to the command line.

This script submits a job in Slurm, in a queue dedicated to visualization sessions. Five servers with 2 L40 GPUs and 768 GB of memory each host the sessions. These GPUs are used by the graphic display, they are not intended to be used for jobs.

Once the job is started, a message is displayed in your command window with the parameters to give to your TurboVNC client on your workstation

Step 2: Connect to the session

On your workstation : launch the TurboVNC viewer client and enter the server and display settings. A second window will ask you for you login and password (same as for SSH command)

A "desktop" window should appear. A red hat in the left upper side corner of the window allows you to open a terminal.

Step 3: Closing the session

Closing the TurboVNC Viewer window does not end the session. As long as the job is running, you can reconnect using the same settings.

To end the session, you must either log out (Upper right corner, on/off sign, then login name and "Log Out") or kill the associated Slurm job (command scancel with the job number).

As for any job, at the end of the duration of the job "TimeLimit" (by default 4 h), the job is killed: the session and all that it contains are also killed.

Default session settings

By default, the following resources are associated:

  • Wallclock: 4 h
  • Number of CPU cores : 4
  • Memory: ¼ of the server, so about 170 GB

If you need more resources, check the startvisu options using the --help argument.


Software, such as Paraview, can be accessed through the module management. Type the command module avail to see the list.

Note: graphics software using OpenGL libraries must be launched via the vglrun command: for example, to launch Paraview, after having loaded the corresponding module (Example: module load /visu/paraview), you must type vglrun paraview.

Amount of memory used

View sessions are Slurm jobs like any other. It is therefore possible to consult the consumption report generated by Slurm.

  • example: sacct -o "MaxVMSize,MaxRSS,AllocGRES,JobID%20,User,Partition" --start 2020-03-23 --end 2020-03-26 --partition visu,visu2

More information about the sacct command in the detailed documentation

Retrieving information from a session

The information of the session is saved in the file ~/.visu/slurm-jobid.out by replacing jobid by the number of the job.

Public IP of the 2 visualization servers:

  • s23visu1 : =
  • s23visu2 : =
  • s23visu3 : =
  • s23visu4 : =
  • s23visu5 : =

Last update: December 4, 2023 10:02:43