advanced rendering for AutoCAD
Tags:
We tried it a while back. Not sure if they've changed anything in the interim but back then we gave up pretty quickly. The basic render farm we sell is file-system based-- that's how the different nodes communicate and how output is extracted. This sort of scheme doesn't tend to run unmodified on a cloud.
The queue management in nXt is very file-system dependent. The servers run all the time as invisible processes-- they place an icon in the tray when running. They periodically access a common folder and sub-folders, watching for certain types of files which cause them to latch onto a task. File locking is used extensively to prevent servers from latching the same task. I doubt if these mechanisms will work properly in ec2. We didn't spend a lot of time on it, though.
At the moment, perhaps. I've got a working prototype cloud solution (works quite nicely actually.) It's scheduled to go into beta testing soon, but I'm somewhat dependent on a 3rd party company for this so I can't completely control the timing.
Will do. This will likely work for path tracer and e4 engines only-- so you may need to experiment a bit with these.
The rate we've been using for testing is about $.30 / core-hr. So, for example, a rendering which would take a 4 core machine 8 hrs. (32 core hrs.) could be processed for about $10.
Make it two customers.
steven talbott said:
If they charge a decent rate per hour like amazon I would be all over this in a heartbeat since my work tends to slam me at odd times and then is dead for a while. Please keep me up to date and let them know they have at least 1 customer ready for this. Thanks!
Nope.