I can’t find a forum dedicated to the IT infrastructure of using Unreal, so given that this regards Lightmass, I’ll put it in Rendering.
If there’s a better sub-forum, I’d appreciate a pointer!
I’m impatient. When baking lighting for levels, I’d love to use temporary computing resources, such as AWS ECC.
In the best of worlds, it would be possible to quickly spin up a fleet of hosts, use a VPN to tie these to my local network, then start the bake, and have Swarm realize that all the resources are available, and bake for me.
Once done, I shut down the resources. Amazon bills per minute, so even using 20 parallel hosts wouldn’t be that expensive (unless I do this all the time, which I don’t.)
However, there are a number of things I don’t know, and need to know in order to make this possible:
- Can a Linux version of the engine run a Swarm agent?
- Does the Swarm agent need a GPU? If so, does it actually do significant work on the GPU, or is a dummy GPU good enough?
- Does each Swarm node need to be connected to the same source control, or will the network transfer all the necessary data?
- Does there exist a Swarm “proxy” that can accept render requests, and when they come in, spin up necessary images and forward the requests to those images? (Well, I can dream, can’t I?)
- How does Swarm agent discovery happen on the local network? Is it through something like UDP broadcast, or do I configure IP addresses in a text file, or what?
I would appreciate any insights into the underbelly of how the IT infrastructure for Swarm rendering works.