@Stephane - any update on scaling since Scale testing LXD clustering?
I am particularly interested in something shallower, meaning more hosts (e.g., 100-1000), less containers per host (1-5).
With less containers per host, I would expect:
- more network traffic on the wire
- smaller spawn rate (longer time to spawn)
Is that your sense?
And, has the container spawn rate improved?
Any other comments would be welcome.