[Ecn-sane] building on BDP

[Ecn-sane] building on BDP

[Dave Taht]

At the stanford buffer-sizing workshop the endless subject of bdp came
up - vs 1bdp and s1/qrt(flows)

In looking over this ..

https://datatracker.ietf.org/meeting/105/materials/slides-105-tsvwg-sessa-61-some-congestion-experienced-00

We're just "used" to a baseline figure in between those and also in no
longer riding the sawtooth on *our* fq'd systems, 'round here.

It struck me how much our codel and fq_codel results (since 2012),
differed from both of theirs and from common industry understandings.
We have amassed so much data over time at so many RTTs that pulling
out the actual BDP's (and how long we have them as "bursts") - might
be useful in terms of restructuring that debate.

Another thing on my todo list is how much BBR and BBRv2 hit fq_codel's
current assumptions. I have a feeling codel is overly aggressive in a
range of rtts (or rather bbr makes a bdp assumption that does not hold
for codel'd networks)

-- 
Make Music, Not War

Dave Täht
CTO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-831-435-0729