This message was deleted.

I don't think of namespaces as a physical thing but more a logical separation. Noisy neighbors etc don't get solved just by namespaces. That's where other things come into play. I'd think autoscaling would be what solves noisy neighbors with the right metrics. Keda for example could be based on a custom metric or pending jobs in a message queue. Just a thought as I'm still picking up kubernetes myself 😉