[Bloat] really lovely live plots of bandwidth, queue depth, marks and drops in libreqos

Dave Taht dave.taht at gmail.com
Fri Jan 6 23:13:55 EST 2023

Wearing my theorist hat... and looking at all the ISP plans
https://payne.taht.net is now emulating.... (click on bandwidth test,
then a plan)

I had never, actually, seen in detail, how cake and fq_codel actually
behave at this range of "plans", starting 2011 with packet captures,
and ns2/ns3 emulations, and mostly working with devices that were
barely capable of pushing 100Mbit, gradually moving up to 1Gbit by
2015 or so. It originally took weeks to get through the packet
captures, then hours to get through and analyze a full range of flent
tests... I'd heard rumors of people pushing cake past 40Gbit, but not
seen anything for myself...

... and now, using herbert's latest rust tooling and his "bifrost" XDP
bridge... and the ebpf pping code with so many contributors... I can
devise a test and watch the results in real time, at a 30ms sample
rate, sufficient, to see slow start, web traffic, voip, and/or other

To see a shaped cake transparent middlebox *pushing* 10Gbits
successfully (the theory should scale, but cpu requirements at the
time I stopped working on cake (about 5 years ago) were too much to
test with) as I just did just now on a mere 16 core box, was really
amazing. Being able to *measure* real 10Gbit traffic while only eating
30% of one of those cores, is also amazing.

Our debloated future looks brighter every day. The (GPLv2) code for
LibreQos - continues to evolve. Please join in at:
https://github.com/LibreQoE - In particular I'd like to find
someone(s) with rust and/or python experience to help out? And does
anyone have 2.5gbit box they could try? I'm pretty sure a pretty
pedestrian, arm box, should be able to run this at 2.5Gbit, if
xdp/ebpf are available on it. Anyone have tests to suggest?

[1] and so far, the only misbehaviors I've spotted are related to a
syn timeout on one of the more gnarly tests

This song goes out to all the folk that thought Stadia would work:
Dave Täht CEO, TekLibre, LLC

More information about the Bloat mailing list