[tsvwg] Re: WGLC for Convergence of Congestion Control from Retained State (draft-ietf-tsvwg-careful-resume)

[Kazuho Oku <[e-mail]>]

Thank you Gorry for the response. +1 to everthing you have written.

2025年3月15日(土) 21:31 Gorry Fairhurst <[e-mail]>:

> On 15/03/2025 17:23, Kazuho Oku wrote:
>
> Hello,
>
> I know it's really late but I finally had a chance to read -15 (the most
> recent version). I see lots of improvements from -12, -15 looks really
> great. Thank you for all the efforts.
>
> When reading -15, I had the following questions floating inside my head. I
> could very well be missing something, or it could be the case that the same
> question has already been asked and answered. If so, please forgive me for
> the noise.
>
> These are good points.
>
>
> Section 3.1:
> > If the measured CWND is less than four times the Initial Window (IW) a
> sender can choose to not save the CC parameters, because the additional
> actions associated with performing Careful Resume for a small CWND would
> not justify its use.
>
> Isn't it two times, assuming that RTT is known during connection
> establishment?
>
> I think the benefit comes when the saved_cwnd is x4, i.e., the jump_cwnd >
> 2 *IW.
>
>
> Section 3.6:
>
> This section talks about RTO, but QUIC does not have RTO. QUIC uses Probe
> Timeouts (PTOs) instead; see RFC 9002 Section 4.7.  Considering that the
> I-D gives specific advices to QUIC implementers in other sections, it might
> make sense to refer to PTO here.
>
> Yes, to make this better suited to QUIC, I suggest:
>
> "A sender that experiences persisent congestion (e.g., a Retransmission
> Time Out, RTO, expiry in TCP) ceases to use Careful Resume."
>
>
> Section 4.1:
>
> > At the end of slow start, the CWND can be significantly larger than
> needed to fully utilise the path (i.e., a CWND overshoot). It is
> inappropriate to use an overshoot in the CWND as a basis for estimating the
> capacity. In most cases, the CWND will converge to a stable value after
> several more RTTs. One mitigation could be to set the saved_cwnd based on
> the flight_size, or an averaged CWND.
>
> I-D suggests the use of flight_size, but doesn’t flight_size share the
> same problem? I think what we might want to suggest here is to use the
> confirmed pipe size (i.e., CWND / 2) if the connection is closed during
> Slow Start.
>
> I suggest:
>
> One mitigation when a connection is in Slow Start could be to set the
> saved_cwnd based on the validated pipe size (i.e., CWND / 2).
>
> Thanks again, do let me know if these changes would address your comments,
>
> Gorry
>
>
> 2025年1月19日(日) 21:25 Marten Seemann <[e-mail]>:
>
>> This email starts a 2 week WG Last Call call to determine if the
>> following TSVWG ID is ready to publish:
>>
>> https://datatracker.ietf.org/doc/draft-ietf-tsvwg-careful-resume/
>>
>> This document targets Proposed Standard.
>>
>> The WGLC will conclude on February 2, 2025.
>>
>> Please do read the draft, and send any comments/concerns to the TSVWG
>> mailing list, including notes on whether these are ready to publish (or
>> send an email directly to the chairs).
>>
>> Best wishes,
>> Gorry and Marten
>> (tsvwg co-chairs)
>>
>
>
> --
> Kazuho Oku
>
>
>

-- 
Kazuho Oku