[Codel] drop state short-circuit...

[Jonathan Morton]

That would imply an unresponsive flow (or a large aggregation of flows in
one queue, which would require a high drop rate to respond).

Indeed, the Codel algorithm in itself doesn't guarantee any latency bound.
Only a hard limit on queue length can do that (or forcibly dropping packets
with sojourn times above the limit).  Practical qdisc implementations do
have such mechanisms.

Flow isolation helps in such circumstances, especially if the qdisc drops
from the longest queue on enqueue overflow.

- Jonathan Morton
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/codel/attachments/20150811/10d01743/attachment-0002.html>