Re: [tsvwg] Prague requirements survey

[Sebastian Moeller <[e-mail]>]

Dear list:


> On Apr 18, 2021, at 12:11, De Schepper, Koen (Nokia - BE/Antwerp) <[e-mail]> wrote:
> [...]
> As a final remark, converging to a steady state rate in the past was always seen as a property of a single mechanism (AIMD of +1 and /2 for Reno, and +cubic(t) and *0.7 for Cubic). I believe we are past simple single ACK response mechanisms (see BBR, ...) where models based on measurements and different states adapt the response and selects appropriate mechanism. When we detect we are out of steady state (0% or 100% marking for a while), the selected mechanism can be RTT dependent (getting up to speed, avoiding latency, ...), once back in sync with the steady marking rate, the RTT independent response can be selected (whatever the mechanism is).

[SM] According to for example https://arxiv.org/pdf/1706.09115.pdf BBR actually suffers from a reverse RTT unfairness, in which longer RTT flows overwhelm shorter RTT flows. While this is intriguingly hinting the possibility of a control loop design that might fall in between the traditional shorter RTT wins and BBR's longer RTT wins, it also demonstrates that RTT-bias is still not a solved problem, even for modern TCP CC as sophisticated and clever as BBR. I take this as support for my hypothesis, that the end-points alone simply are not well suited to solve this problem; no matter how convenient that would be.


Best Regards
	Sebastian