Is it possible to use a smaller instance type for the infrastructure? At the very least for testing a 2XL seems excessive. I can stop it and change the type and resume it but that’s a bit awkward.
No, but I’ll make it a feature request here. If anyone else wants to be updated on that, feel free to post here and I’ll add you to the dev issue (which is my usual protocol).
I think a big issue at the moment is that you may not know when the instance becomes to bottleneck. We’ll have to find a way to make that visible too, so I’ll go ask that of the dev team too.
Any updates here? I just did a monthly estimate and AWS said $500 a month. LOL No. Not paying $500 a month for cloud rendering to be available.
As a benchmark a C4 Large can handle about 50-60 render nodes. So a 2XL should only be necessary (without any performance impacts) with >200 render nodes. Considering I had to petition AWS to even allow me to have 50 spot instances (and even then I still get throttled) and a C5 18XL is a beastly machine I would have a couple buttons like you have on spot fleets that have some rough numbers:
Small: 10-50, Medium: 50-100, Large: 100-200, XL 200-500 render nodes.
If you need more than 500 18XLs… you’re probably wanting more control over the infrastructure anyway and need a more custom solution.
Correction. I’ve got 30 AWS nodes running right now on a T3 Medium and I’m still gaining significant CPU burst credits I bet I could sustain at least 50-60 on a T3 medium (with a burst when new spots fleets startup). That’s only $0.04 an hr/ vs $0.40/hr for the XL default. And with Unlimited mode enabled it will only charge extra when rendering is heavy.