[kryptk]

What about resource allocation? With Kubernetes I can specify both requests and limits of cpu and memory and it will fill my nodes to match. A node autoscaler gets me more nodes if needed. I can define vertical pod autoscaler that can dynamically modify those requests and get me an emptier or bigger node if needed. I can define a horizontal pod autoscaler to keep aggregate cpu at a set target by auto-spawning more containers for me, and k8s handles load balancing via dns. Do typical production setups not require some/most of these features? Mine do.

[asdfaoeu]

Except for the vertical scaling all of that can be done with asgs/elbs.