We do use a type of queuing but not CBQ or Fair. Both add a high degree of latency to every packet because it has to be queued. We have a weighted priority scheme that performs the dual task of sending critical data first as well as limiting drops of less critical data. We call that our anti-chocking technology. We also apply priority thresholds into order to reduce latencies when the network usage is below a certain percentage because the best thing that we can do in those cases is stay out of the way.

We start prioritizing when network load gets high enough to make it necessary and the small latency that we add then is inconsequential compared to drops or discards. Our technology also works in both directions while CBQ and Fair queuing can only work on traffic outbound on an interface.