Answer the following questions:

• What is the user experience like when troubleshooting when one node on the data plane is down?
• What is the user experience like when troubleshooting when a service on a data plane node is broken in some weird way (eg: perhaps libvirt will not come up)?
• How long does the execution of the data plane work take? Is it linear, based on the number of nodes or does it scale in a different way?
  • Initial bootstrap/install for the whole data plane
  • Re-run bootstrap/install after resolving a failure on a subset of data plane hosts
  • Implement a change in configuration which applies to the whole data plane (test one that implements via nova operator, and one via data plane operator)

Come up with suggestions for improvement to find a good mix between execution time and usability.