carpatic wrote:
> I want to achieve the following: I have a Flash application for doing online
> some design/modeling, and I want to render the resulting scene with povray
> (I'm taking the xml output from Flash, generate a pov file through a python
> script and call megapov to render the scene), then I load the outputed
> image in the browser.
> 
> My question is the following: which is the hardware recomandation for the
> machine that is doing the randering, in order to ensure a good raport on
> speed/quality ? Let's consider the example of 10 symultaneous randerings
> for 800x600 px, top quality, fairly simple objects (mostly boxes) and
> simple textures.
> As it runs over the web and the user expects rapid results on his screen,
> efficient timing is the most important factor to be achieved.
> The hardware requirements are for a dedicated server, no clustering, no LAN
> randering distribution.
> 
> Thanks
> 
> 
> 

Hmm, for a single dedicated server, a X1E Cray should fit the bill 
nicely... 8^)

I built a "real time" raytraced imaging system about a year ago as a 
proof of concept design, to do basically what you're describing. 
Raytracing and rendering objects in the background and then making the 
completed image available to the user. It worked ok with Dual 2.8 Xeon 
processors running SuSE Linux, 4 GB RAM and a SCSI array. The problem is 
in spawning several instances of povray to handle the rendering. Even 
with a dual Xeon, it can eat up resources.

Consider having serveral users all requesting raytracings 
simulteneously... not practical.

After playing with it a bit I decided to take a different approach.

I'm now using the MOSIX kernal patch and some fancy programming using 
ERLANG to distribute my renderings between several servers. It smokes...

Post a reply to this message