Hello David
Nice to see you after a while!
Yes that is correct. Queuing only takes place when there is congestion. Think about it this way. If you go to a bank, and there is no lineup, you are served immediately. There is no queue to manage!! So there is no meaning to employing CBWFQ/LLQ or any other queue management algorithm if there is no queue!
Ah yes. The question here is “what is congestion?” Is it when the port itself reaches its capacity or does congestion also occur when the CPU and memory resources of the network device are overwhelmed? Or is it when the backplane of the device has reached capacity, even if the specific port in question has not?
QoS mechanisms are primarily designed to deal directly with port-level congestion. When the port’s egress queue (or ingress queue, depending on the device and QoS design) fills up, QoS mechanisms like traffic shaping, policing, scheduling, and queue management kick in.
So to answer your question, congestion, as it pertains to QoS, is most commonly tied to oversubscription of physical or logical interfaces and queue overflow. CPU, memory, or backplane congestion can also exist but are not the typical focus of QoS mechanisms.
As always, it’s a pleasure to answer your questions…
I hope this has been helpful!
Laz