Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
There could be several reasons for you AF11 tail drops. WRED uses an average, so bursting can go beyond what's desired as packets are not dropped soon enough to avoid tail drop. A variation of this problem, WRED drops packets as they are added to the queue, so here too, flows might not see the drops soon enough to avoid bursting beyond desired limits. The forgoing issues are compounded when you use large queue limits.
Another possible issue, is all your AF traffic drop responsive? Non-drop responsive traffic will not moderate their transmission rates which can cause tail drops.
Assuming the AF class is also deprioritzed relative to other traffic classes, a sudden decrease in class available bandwidth can cause immediate queuing congestion; again so quickly, the early drops do not have sufficient time to have the transmission sources slow their transmission rates.
WRED can be very difficult to tune for optimal results.
