[Cake] AICP

[Jonathan Morton]

> On 4 Aug, 2018, at 11:11 am, Dave Taht <dave.taht at gmail.com> wrote:
> 
> https://sci-hub.tw/10.1109/CNSM.2014.7014160

Looks like an advance over the status quo (as illustrated by Google's paper) with respect to policers.  The basic TBF-type policer is pretty crap; really not hard to do better if you bother to think about it.

I'm still not *entirely* convinced that we can't have real DRR++/Codel implementations at line rate in hardware. I think nobody has even seriously tried to do it.  Getting to that point would potentially obsolete policers overnight, at least from a theoretical point of view.

Meanwhile, we can still do a lot at the gigabit-per-subscriber level in software.

 - Jonathan Morton