[GSoC 2012] Clustering RenderSystem

[Nodrev]

Hello,

My idea is related to something I often ear while on duty, and it's concerning VR usage for Ogre. I propose to add an OpenGL derivated render system (derivating the OpenGL plugin), that could manage clustering rendering.
There's already the Chromium/DMX solution as discussed on the wiki, but it's not what I have in thoughts.
I think about integrating Equalizer library directly in Ogre. This way, for example, we could launch an existing ogre application with an 4 windows configuration load balanced, or launch a master application on a server and the same application as slaves on multiples computers!
The base idea is to have a custom RenderSystem (but maybe other components are needed) that will handle by himself the cluster.
Of course, this need more deep think, but I just want a rapid opinion of it's feasibility and if it's well suited for an GSoC

[Nodrev]

Arg, didn't saw that the studient deadline was tomorrow... probably too late for such a proposal, anyway, your opinions on the question interest me

[duststorm]

I think it's a very interesting proposal. One that could attract a new userbase to Ogre.
I personally don't have much experience in this area, but I know Ogre is used in a lot of serious OpenGL targeting applications such as simulation and visualisation. In any of the cases where distributed rendering is needed Ogre falls out of scope at the moment (unless one decides to do a lot of customization).

So yes, very good project idea, and it fits well for being an Ogre core feature.

[alfski]

Hi - I'm from the Liquid Galaxy GSoC Project - http://www.google-melange.com/gsoc/org/ ... uid_galaxy

We had a student last year working on immersive on a clustered GL wrapper for generic Linux OpenGL applications..

Paul's project is here - http://code.google.com/p/clustergl2/

The idea is to to run 'normal' OpenGL apps from a master machine and have the output displayed immersively/tiled onto Liquid Galaxy rigs.

BTW we'd be happy to be guinea pigs for any GSoC students working on networked rendering for Ogre.

Cheers, Andrew
eResearch / Uni of Western Sydney

[Nodrev]

I think it's a very interesting proposal. One that could attract a new userbase to Ogre.

Indeed, I recently talk to lots of students and searchers that, even if they knows Ogre, will never use it because they need clustering (or at least multiscreen, which is from my point of view barely supported by Ogre, on the paper everything is right, but just try to change from one display to another a render window, or use non fixed resolution size, and the problems will start...).

Hi - I'm from the Liquid Galaxy GSoC Project - http://www.google-melange.com/gsoc/org/ ... uid_galaxy
We had a student last year working on immersive on a clustered GL wrapper for generic Linux OpenGL applications..

Hi alfski, welcome! I already know this project, but I didn't talk of it in my first post as there's one big problem with it for being an Ogre dependency: it's not multiplatform, it only works on linux, and furthermore, it have an sdl dependency (if I remember well) which was for an unremainded reason removed from Ogre's OpenGL render system some years ago.
It's a great project, working on OpenGL instruction interception (as Chromium), which is different from Equalizer. But I only have small experience on the question, and the advantages of using one or other solution are not fully mastered by me at this time
The other difference with Equalizer is that it seem (but maybe I'm wrong) that your solution only works for a "wall layout", Equalizer is far more flexible on this question: http://www.equalizergraphics.com/scalability.html

BTW we'd be happy to be guinea pigs for any GSoC students working on networked rendering for Ogre.

Understood!

[spacegaier]

Arg, didn't saw that the studient deadline was tomorrow... probably too late for such a proposal, anyway, your opinions on the question interest me

Not necessarily too late, since today is only the deadline to enter the proposal on Google Melange. It is not yet the day that a decision about the GSoC will be made, so you might i fact still argue a strong case why your project should be considered.