[RC5] film rendering

Michael L. Dawson mdawson at ix.netcom.com
Sat Jan 23 14:01:29 EST 1999


re: film rendering

There's a lot more to this than meets the eye. First off, there are already
graphics programs that will let you install a copy of a licensed copy on any
number of machines strictly for network rendering purposes. These programs
already have their render engine built-in and the host copy takes care of
dividing up the load across the network. This is proven technology. All that is
required is usually the licensed copy and multiple machines on a network. Since
those programs operate on a per frame basis, it's senseless for them to go to a
per pixel basis of load distribution, but it might work for a d.net style
render farm........ but then you're going to add a lot of overhead just to
disseminate and reassemble the pixels.

Still, there are other aspects to consider. There won't be just the final
render at high resolution. There will be countless revisions and re-renders at
lo-rez just to make sure the film is progressing on track. This kind of load
over time can be as much, if not more of a load than the final render itself.
Then there needs to be a way for the film client to see any work as soon as it
is done. The closer to real time that is, the better chances of getting
something like this to fly. Looking into my crystal ball, I can see where
somebody like d.net would want to contract for spare cycles just so that they
could accurately predict response times for the film client!

Then there's the delivery. That rendered data has got to be fed to a film
recorder. Relying on an Internet feed for a data source is tantamount to
pissing in the wind, in my humble opinion. Totally unreliable for an
application such as this, unless the data could be stored onsite by whomever
owns the film recorder. Somebody like that may or may not have the terabytes of
hard disk space necessary for something like this. One way to circumvent that
would be to use high definition videotape as an interim storage medium, and
that could get a bit costly.

So maybe we shouldn't start with a film. How about a half hour show, like South
Park or The Simpson's, a 5 min cartoon, or even a lowly :30 PSA about
distributed computing aimed for the television medium rather than film. The
data rates and time factors are much lower and might work out to be a good
proving ground for the day we do get to render a feature length film.

but it would be fun.....

mdawson

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list