Q & A. Learn more about the the gfxlauncher (SLURM interactive application launcher)

It is interesting to learn more about the ThinLinc ecosystem as it allows us to understand the use of ThinLinc for diverse purposes.

This is the case with the gfxlauncher, for instance.

Below I share some Q&A we had about the gfxlauncher. The questions are answered by Anders Follin and @jonas_lindemann from Lunarc.

Why was gfxlauncher created?

The main reason for the development of gfxlauncher was to manage resource allocation through SLURM in an easy to use and fair way instead of using one or more static graphics servers.

What is the vision for the project?

The vision is 3-fold:

  1. Provide an easy to use way for users to specify resource needs for interactive applications (GUI/OpenGL/Notebooks).
  2. Enable more user groups to take advantage of HPC resources.
  3. Enable more efficient resource sharing in an HPC environment, reducing the load of the login-nodes by offloading intractive applications to dedicated resources.

How does it work together with ThinLinc?

LUNARC is using Thinlic as the supported, scalable and performing “2D pixel transport” vehicle for our infrastructure. We can focus on and develop the backend services with purpose to enable enhanced scientific workflows including accelerated 3D graphics, multiple OS’es, etc.

What kind of organizations/ users would benefit of gfxlauncher?

  • Organisations with a Thinlinc enabled HPC environment.
  • Organisations that wish to more easily and effectively share resources for interactive applications.

What kind of applications are commonly used with gfxlauncher?

PhotoScan, MetaShape, ParaView, ABAQUS/CAE, FLUENT, ParaView, …

Why is better to use gfxlauncher when compared to other methods?

Provides a very configurable application launcher that can an easy to user interface for resource selection and launch options. Currently support launching standard X11 applications, Python-notebooks, MATLAB-sessions, hardware accelerated applications and Windows based sessions through SLRUM/RDP.