I was going to go into a sidebar in another thread [I nearly did] but I wanted to talk about CLOUD and DEADLINE and what we did at the show.
for months during internal alpha builds I was managing the farm and test-farm from Los Angeles - from the office, from my house etc. I pushed really hard for this to feel performant, the guys had a lot of ideas which we implemented [background threads, ui threads, managing the data throughput] - we tested a variety of things large and small from loading of icons, to how quickly you can work in the ui while data loads in the background etc. In the end, D6 was extremely functional in a remote environment.
We then tested with the database on EC2, slaves on EC2, database local etc - and a variety of combinations. In the end, we found D6 to be stable, reliable and allow the flexibility of combining remote machines into one coherent, logical slave pool.
Once we accepted that, we decided the coolest demo would be to have a LIVE demonstration of rendering across physical locations, [and actually doing it] and with the caveat that ‘oh it isn’t going to finish in time’ - be able to SCALE live at the show to get the job done by the end of the presentation.
We realized we had all of the same challenges a facility would have - how to synchronise the actual scene data, the renders etc. That’s an enormous task to build a pipeline to do that in a few weeks - but since we had recently adopted Dropbox internally as a way to synch sample videos, scene data and other things for Siggraph - we tried it out on our remote rendering. And it worked. It worked very well.
Now, i’m not going to suggest that this is the final solution - but for small facilities, or specific jobs- it very well may be a working solution, TODAY.
more to come…
cb