Date: Wed, 26 Oct 2022 15:03:04 +0200 From: tuexen@freebsd.org To: Tom Jones <thj@freebsd.org> Cc: Hans Petter Selasky <hps@selasky.org>, Zhenlei Huang <zlei.huang@gmail.com>, freebsd-net@freebsd.org Subject: Re: Too aggressive TCP ACKs Message-ID: <701FC3D6-7AFE-46F6-977D-75CAD34646D4@freebsd.org> In-Reply-To: <Y1kvNBMedJtrs6yo@spacemonster> References: <75D35F36-7759-4168-ADBA-C2414F5B53BC@gmail.com> <712641B3-5196-40CC-9B64-04637F16F649@lurchi.franken.de> <62A0DD30-B3ED-48BE-9C01-146487599092@gmail.com> <0FED34A9-D093-442A-83B7-08C06D11F8B5@lurchi.franken.de> <330A9146-F7CC-4CAB-9003-2F90B872AC3E@gmail.com> <1ed66217-5463-fd4d-7e7a-58d9981bc44c@selasky.org> <Y1j2ZzHaFt/YA5Et@spacemonster> <4E92E238-798B-4293-B0D2-81E3FCB92E34@freebsd.org> <Y1kvNBMedJtrs6yo@spacemonster>
next in thread | previous in thread | raw e-mail | index | archive | help
> On 26. Oct 2022, at 14:59, Tom Jones <thj@freebsd.org> wrote: >=20 > On Wed, Oct 26, 2022 at 02:55:21PM +0200, tuexen@freebsd.org wrote: >>> On 26. Oct 2022, at 10:57, Tom Jones <thj@freebsd.org> wrote: >>>=20 >>> On Sat, Oct 22, 2022 at 12:14:25PM +0200, Hans Petter Selasky wrote: >>>> Hi, >>>>=20 >>>> Some thoughts about this topic. >>>>=20 >>>> Delaying ACKs means loss of performance when using Gigabit TCP=20 >>>> connections in data centers. There it is important to ACK the data = as=20 >>>> quick as possible, to avoid running out of TCP window space. = Thinking=20 >>>> about TCP connections at 30 GBit/s and above! >>>>=20 >>>> I think the implementation should be exactly like it is. >>>>=20 >>>> There is a software LRO in FreeBSD to coalesce the ACKs before they = hit=20 >>>> the network stack, so there are no real problems there. >>>>=20 >>>=20 >>> Changing the ACK ratio seems to be okay in most cases, a paper I = wrote >>> about this was published this week: >>>=20 >>> https://onlinelibrary.wiley.com/doi/10.1002/sat.1466 >>>=20 >>> It focuses on QUIC, but congestion control dynamics don't change = with >>> the protocol. You should be able to read there, but if not I'm happy = to >>> send anyone a pdf. >> Is QUIC using an L=3D2 for ABC? >=20 > I think that is the rfc recommendation, actual deployed reality is = more > scattershot. Wouldn't that be relevant? If you get an ack for, let's say 8 packets, = you would only increment (in slow start) the cwnd by 2 packets, not 8? Best regards Michael >=20 > - Tom >=20
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?701FC3D6-7AFE-46F6-977D-75CAD34646D4>