Thank you Jonathan. I'm going over the code to learn more about CAKE. This
set-associative hashing seems to be pretty useful. I'll ask more questions
about the code as I try to understand it better.

On Sun, Oct 29, 2017 at 12:14 PM, Jonathan Morton <chromatix99@gmail.com>
wrote:

> Yes, in the same way as is caused by SFQ's periodic hash perturbation.
>
> However, in Cake this is a rare condition brought on by exceptional load;
> it typically takes hundreds of bulk flows to get an unresolvable
> collision.  The use of AQM also means that the extent of reordering is
> reduced due to shorter average queues.  Most sane protocols, including
> TCP-SACK, should cope fine.
>
> - Jonathan Morton
>