Maybe suggest using BBR congestion control?

[ivanbaldo]

For kernel >= 4.9 maybe it should be suggested to add this in sysctl:
net.ipv4.tcp_congestion_control = bbr
And this in /etc/modules:
tcp_bbr
Maybe I am wrong though... but if thats the case, then it could be mentioned in the webpage why I am wrong so others don't ask themselves this same question.
Thanks a lot!

[dtaht]

I don't regard BBR as baked enough to recommend.

[ivanbaldo]

Thanks a lot for your answer Dave!
Now more than two and a half years later, whats your opinion on using BBR?
BTW I been using it together with fq_codel and ECN=1 since then at home and one of my customers, didn't notice any problems so far... (of course this is just my case and I am not as expert as you, just comenting).

[dtaht]

Iván Baldo <notifications@github.com> writes:

Thanks a lot for your answer Dave! Now more than two and a half years later, whats your opinion on using BBR? BTW I been using it together with fq_codel and ECN=1 since then at home and one of my customers, didn't notice any problems so far... (of course this is just my case and I am not as expert as you, just comenting).

Ultimately, ecn or no, BBR will construct a good model of the path, and fq_codel garuntees that that is, indeed a good model. That said, I'd really like to see RFC3168-like behavior enter bbr.

…

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.