Why pacing gain of 0.9x is better for fairness than 0.75x in down phase?

144 views
Skip to first unread message

ssw

unread,
Apr 8, 2024, 2:43:09 AMApr 8
to BBR Development

Hello! I was reading changes of BBR v3 in IETF 117. And i got a question.

In bbr v3, pacing gain of down phase was changed to 0.9x. But i could not understand why pacing gain of 0.9x is better than 0.75x.

Because as slow flows decrease pacing gain to 0.75x, faster flows will do same thing. So, I thought if range of sending rate was wide, it would help getting fairness. But pacing gain was changed to 0.9x in v3.

Could you help me to understand this change?

I appreciate you for using your time to read this question!

Best regards,
Seoungwon

cutesocks

unread,
Apr 17, 2024, 11:48:33 AMApr 17
to BBR Development
Screenshot from 2024-04-17 21-57-48.png
my guess is just as the ppt said, pacing gain 0.9x and 2.25x cwnd gain that allows slow flows to consistently utilize enough bw to cause fast flows to yield bw for fairness convergence, simply just better than the old param.
Reply all
Reply to author
Forward
0 new messages