Material Database: Server should render previews

[psor]

It would be really good to have the server rendering the previews like it is with the luxrender matDB.
You'll find out very soon, that a lot of materials are not getting uploaded because of not spending the
time to let render to a fully cleaned preview image. Ask Dough Hammond he implemented the stuff
for luxrender and is also known as dougal2 on this forum.


Please think about it. I'm taking a guess that more people will upload materials then. ;o))






take care
psor

[Doug Armand]

Actually the reason I don't upload is because most of the materials I create are for a specific shot and are intended to be rendered from a certain distance/viewpoint and therefore probably wouldn't be very good quality for general use.

But that's just my reason

[psor]

I've another idea. Nick could integrate a function into the material editor to render a list
of materials. That would at least solve the issue with "Aaargh, I have to do it manually".


Of course it would be better to render with the server as it would save us money, time
and would use a system that is running 24/7/365 anyway.


And if somebody shouts, you can write a script for batch rendering your self, then I
can only answer. Just do it! Then do it proper and do it for linux/mac/windows. ;o))





take care
psor

[pixie]

+1

[PureSpider]

I talked about that idea with fused a loooooong time ago, back when we still had his mat db... ^^
I think its a nice idea and I would support it!
Writing a script for rendering it on the server is not really hard, anyway.

[fused]

Yeah, back then, when i wrote the old matDB code, talked to Nick about it and he didn't really like the idea.

But then agin its nearly 2 years ago now. I still like the idea

[dougal2]

The luxrender lrmdb implementation of rendering has 2 main elements:
1. The database system has an XMLRPC server built in, so that any part of the system can be executed remotely.
2. A python daemon that connects to the XMLRPC service, logs in, downloads data, renders it and uploads the result image.

The lrmdb system is GPL licenced, it would be no problem (and also a great honour) to see the code/architecture re-used.

I have found however, that trying to run the daemon on the webserver was problematic. At times it visibly affected the website loading times, at others it just crashed inexplicably.

My original intention was that anybody could download and run the daemon script in order to donate some rendering time. This also never really happened - partly due to trust and security concerns, and partly because the lrmdb doesn't really require that much computing power.

Recently I have also constructed a linux VM image that could be distributed for anyone to run (similar to the vswarm idea). Again though, there's no point because of the lack of demand.

The solution that lux has at the minute is that I occasionally run the daemon manually myself at home when there's some rendering to be done. Even an hour or so run-time is enough to bring new materials online and to add a few samples to a few others.
Unfortunately, the lrmdb doesn't get used very much - I run the daemon perhaps once every 2 - 3 weeks at the minute.

I'd be happy to share implementation ideas with you if this is the route you guys want to go with the indigo matdb.

[OnoSendai]

There's no way we can render the previews on the server, it's loaded enough as it is.
Dougal2's idea is interesting, but I don't think it's worth the hassle to set up.