[nurble]

I’m working at a shop now that has 8 burn servers being shared by a dozen or so flames. it’s working so well that there’s already talk of getting another bunch (12 or so) to round out the farm. there are bottlenecks, to be sure. it’s pretty difficult to share servers amongst multiple jobs, so we end up waiting in line for our renders to pop out, but it’s still a massive timesaver. we’re working 2k 12bit and the render times are faster on a single burn server than on the octane 2. sparks scream through in probably half the time of a local render, and that’s just with one node! So when you get 8 dual proc machines all chugging away at your clip, something that would have taken 4 hours on the octane can be finished in ten or fifteen minutes with burn. it’s pretty amazing.

there are lots of issues still, memory and bandwidth problems account for most of them; large setups end up with dropped frames or skipped nodes in batch. with things like regrain rarely working at all. the backburner software isn’t too intelligent about sharing resources, as i said, so you have a lot of “can i jump ahead of you in line” type conversations happening between artists. But even through all of that, i don’t think we’d ever consider going back to the old way. it’s certainly changed the way i work, and the next facility I stroll into will get an earful about how great it is.

i wasn’t around for the setup, but i’m told it wasn’t pretty. you have to know about linux to get the burn servers running, and you need to know windows to get the backburner server running, and there are wire and spark issues to be dealt with (and licensing all those sparks isn’t cheap). is it all worth it? from a productivity standpoint, absolutely. One flame with 4 burn servers would see an exponential increase in productivity, particularly for people working on more than one thing at a time.

[Author]

Posts