Re: [tcpm] I-D Action: draft-ietf-tcpm-hystartplusplus-12.txt

[Neal Cardwell <[e-mail]>]

Your proposal to remove the description of the Linux TCP experience from
the "Deployments and Performance Evaluations" section sounds very
reasonable to me. I don't have any concerns with removing that completely.
I agree that it mostly just seems to cause confusion.

thanks,
neal


On Wed, Jan 11, 2023 at 6:47 PM Praveen Balasubramanian <
[e-mail]> wrote:

> I see, thanks Neal for the clarification. I then suggest that we eliminate
> this deployment experience line completely because its not with the
> proposed algorithm in this document and that it will confuse new developers
> as to why the document suggests using pacing with L=infinity when the
> default in Linux contradicts it.
>
> Any concerns with just removing this line entirely?
>
> On Wed, Jan 11, 2023 at 2:07 PM Neal Cardwell <[e-mail]>
> wrote:
>
>> Martin pointed out out-of-band that my various attempts at explaining
>> this point have not been clear, so let me try to express this a different
>> way:
>>
>> For about the last decade the Linux TCP default has been: [ CUBIC +
>> Hystart + L=infinity + unpaced ]. This is because the commonly deployed
>> qdiscs for the major Linux distributions do not implement pacing.
>>
>> For 2013-2016 Google/YouTube Linux TCP ran with: [ CUBIC + Hystart +
>> L=infinity + paced ]. This was using the fq qdisc to implement pacing.
>>
>> best regards,
>> neal
>>
>>
>> On Tue, Jan 10, 2023 at 10:24 AM Neal Cardwell <[e-mail]>
>> wrote:
>>
>>> Looks like there is still one significant word in the "Deployments and
>>> Performance Evaluations" section that needs to be updated to be accurate.
>>>
>>> Existing draft-ietf-tcpm-hystartplusplus-12 text:
>>>
>>>   There has been over a decade of experience using the
>>>   original Hystart algorithm for all TCP connections in the
>>>   Linux operating system with pacing *enabled* and an
>>>   actual L = infinity.
>>>
>>> Suggested text:
>>>
>>>   There has been over a decade of experience using the
>>>   original Hystart algorithm for all TCP connections in the
>>>   Linux operating system with pacing *disabled* and an
>>>   actual L = infinity.
>>>
>>> I mentioned the rationale in the "Re: [tcpm] Changes to
>>> draft-ietf-tcpm-hystartplusplus after WGLC" thread on Nov 29:
>>>
>>>
>>> https://mailarchive.ietf.org/arch/msg/tcpm/CSzCgxxSLrO9u63dymyyxqB6bD4/
>>>
>>> cheers,
>>> neal
>>>
>>>
>>> On Mon, Jan 9, 2023 at 8:53 PM <[e-mail]> wrote:
>>>
>>>>
>>>> A New Internet-Draft is available from the on-line Internet-Drafts
>>>> directories.
>>>> This draft is a work item of the TCP Maintenance and Minor Extensions
>>>> WG of the IETF.
>>>>
>>>>         Title           : HyStart++: Modified Slow Start for TCP
>>>>         Authors         : Praveen Balasubramanian
>>>>                           Yi Huang
>>>>                           Matt Olson
>>>>   Filename        : draft-ietf-tcpm-hystartplusplus-12.txt
>>>>   Pages           : 9
>>>>   Date            : 2023-01-09
>>>>
>>>> Abstract:
>>>>    This document describes HyStart++, a simple modification to the slow
>>>>    start phase of congestion control algorithms.  Traditional slow start
>>>>    can overshoot the ideal send rate in many cases, causing high packet
>>>>    loss and poor performance.  HyStart++ uses a delay increase heuristic
>>>>    to find an exit point before possible overshoot.  It also adds a
>>>>    mitigation to prevent jitter from causing premature slow start exit.
>>>>
>>>>
>>>> The IETF datatracker status page for this draft is:
>>>> https://datatracker.ietf.org/doc/draft-ietf-tcpm-hystartplusplus/
>>>>
>>>> There is also an htmlized version available at:
>>>> https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-hystartplusplus-12
>>>>
>>>> A diff from the previous version is available at:
>>>>
>>>> https://author-tools.ietf.org/iddiff?url2=draft-ietf-tcpm-hystartplusplus-12
>>>>
>>>>
>>>> Internet-Drafts are also available by rsync at rsync.ietf.org:
>>>> :internet-drafts
>>>>
>>>>
>>>> _______________________________________________
>>>> tcpm mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/tcpm
>>>>
>>>