Hey, sorry I didn't see this question before: Meraki switches only apply the queuing and scheduling part of QoS. So no token bucket policing here 😉 Also which is a shame, there is no strict PQ. All Meraki switches basically have 8 shared round robin bandwidth queues where each following queue has double the bandwith guarantee than the one before it. But if said queue is not fully utilized it will yield the remaining bandwidth to the next lower queue. Queue 6 and 7 are reserved for internetwork and network protocols so they have the biggest allocations and queue 0 - 5 are configurable (they call it CoS queues). So you can map DSCP EF + AF41 to queue 5 if you want the max bw for those. You will find however that most examples only use up to queue 3 since they don't really use that 12 tier model Cisco uses to translate into switches. So all traffic that does not get scheduled remains in the queue in memory buffer or tail dropped if the memory buffer is full. There are no advanced tuning methods like RED/WRED or policing to avoid said congestions.
... View more
