Hi,
I’m increasingly having internal discussions about the use of cloud based solutions vs buying more physical rendernodes for short-term, heavy ramp-ups in our pipeline. Just saw this this morning:
cgarchitect.com/news/newsfeed.asp?nid=5303
It would be good to get some insight from Thinkbox on your current line of thinking on this topic! Can we expect to see Deadline in a cloud based environment any time soon?
Thanks,
Mike
July 18, 2011 | Mike Owen
I am very interested in such topic as well as i plan to set up a few cloud based nodes at some point in the future (near or far). Utilizing unused CPU power for cheap seams to be a future model that could work.
July 18, 2011 | Anselm v. Seherr - Thoss
I dont know how to answer this - we have something in development. i hope to be able to show off at least a tech demo of it at siggraph, but with all of the other things we are showing off, i’m ot sure it will make it. that being said, if we decide not to show it off at siggraph - we’ll show it off after siggraph.
if a facility or project needs something, i’m sure we can provide a custom solution - but for a specific general-purpose tool: that will be a bit longer.
does this help?
cb
July 19, 2011 | Thinkbox Software
If Deadline could manage VM’s, you could render on cloud servers like EC2, but you could also render on my servers. If there was a marketplace where you could buy/sell/put/call on render resources we could achieve fairly good efficiencies ourselves. I suspect we would even be cheaper than EC2.
The problem with any cloud system will be reducing bandwidth, which is MUCH more expensive than the CPU cycles. You don’t want to package up everything needed to render, just the essentials, and that would mean diffs or similar. Maybe even just sending a compressed packet of instructions to a running VM. I’m no expert on cloud deployment, but the way we’re used to managing a farm does us no good here.
July 19, 2011 | Chad Capeland
Thanks for the insight Chris.
I think technologies such as Aspera are going to be important in the future to help shift data around using existing bandwidths. For example, due to the geographic location of our office, we have a fixed limit on how fast we can ever go without having to bring in another dedicated line. However, by using Aspera we can get much more out of our existing connectivity.
Just like Chad said, for us, it’s also about how to shift the min. amount of data around to make it all work.
Mike
July 19, 2011 | Mike Owen
Chad,
my roadmap is for us to manage vm’s. i’m nearly at the point of putting some preorders up a la kickstarter to test some of these ideas in the marketplace and see who would buy it 
mike: aspera, we’ve also got a roadmap for our own udp transfer system. aspera is pricey.
cb
July 19, 2011 | Thinkbox Software
Just a correction, EC2 apparently doesn’t charge for data upload, only data download. So assuming you are only getting a few TB of .exr files back, it’s not that bad. Managing redundant uploads to save time (so we don’t send the same 60GB of PRT’s each time we need to tweak the shadows) would still be something useful to have though.
Interestingly, the machines on EC2 aren’t that impressive (at least by VFX standards). Equivalent to a high end i7 or e5. So the benefit isn’t speed of the machines, but speed of buildout and setup.
- Chad
July 20, 2011 | Chad Capeland
Just flew over the thread so far! Sounds all great 
Can’t wait for the VM management system Thinkbox is coming up with. I will have a few test VMs at the university servers here by the end of this week! They will buy deadline and Krak then but prolly will email first for an eval license.
I will mail you separte, cb!
Ansi
July 20, 2011 | Anselm v. Seherr - Thoss