understanding model in a cluster

[jochen-schuettler]

Please tell me if my understanding is correct:

• in a VM kepler identifies CPU and RAM of the hardware machine it's running on
• it uses a model of a machine with identical setup to evaluate the consumption
• a K8s cluster consists of nodes, which might run in the same VM or elsewhere
• so the identified CPU and RAM might not fit to all nodes, meaning the model is off by that degree
  I'm clear about the problem that consumption curves are nonlinear, and we don't know what else is running on the same server with our VM and that might be a bigger error factor. Still my question remains.