[Cake] STEAM tcp algo from CDN?
Taraldsen Erik
erik.taraldsen at telenor.no
Wed Mar 10 02:04:02 EST 2021
So to summarize, not quite certain which tcp algo they use. But everyone seems to agree that Steam does whatever they can to jam the pipe full
________________________________
Fra: Dave Taht <dave.taht at gmail.com>
Sendt: onsdag 10. mars 2021 01.57.09
Til: Nils Andreas Svee
Kopi: Jonathan Morton; Toke Høiland-Jørgensen via Cake; Taraldsen Erik
Emne: Re: [Cake] STEAM tcp algo from CDN?
I would really like them to use less flows - or BBR started up
separately with a 200+ms
interval between starts would be good....
On Tue, Mar 9, 2021 at 4:31 PM Nils Andreas Svee <me at lochnair.net> wrote:
>
> On Wed, 2021-03-10 at 01:14 +0200, Jonathan Morton wrote:
> > > On 9 Mar, 2021, at 10:20 pm, Dave Taht <dave.taht at gmail.com> wrote:
> > >
> > > 10-20 flows, cubic, last I looked. It's ugly.
> >
> > I can't confirm CUBIC from here, but it seems to be 4-8 flows in
> > parallel now. Latency to the national CDN is about 22ms over LTE, so
> > it's hard to distinguish CUBIC from anything else in particular; in
> > this range it would look a lot like NewReno. It seems to shut down
> > each flow and start a fresh one after about a minute.
> Seeing somewhere between 8-16 here I believe. IIRC I've seen you
> mention being located in Finland, if so we're probably on the same
> datacenter in Stockholm.
>
> The flows are probably not lasting long because the downloads are split
> into chunks, which according to my logs are at most 1 MB. Patches seem
> to work differently, those can be much larger (logs shows the largest
> one the past months was 137 MB, depends on the game I assume).
>
> --
> Best Regards,
> Nils
>
--
"For a successful technology, reality must take precedence over public
relations, for Mother Nature cannot be fooled" - Richard Feynman
dave at taht.net <Dave Täht> CTO, TekLibre, LLC Tel: 1-831-435-0729
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/cake/attachments/20210310/24ced7e1/attachment.html>
More information about the Cake
mailing list