On Fri, Jul 9, 2021 at 4:07 PM Daniel AJ Sokolov <daniel@sokolov.eu.org> wrote:
>
> On 2021-07-09 at 3:58 p.m., David Lang wrote:
> > IIRC, the definition of 'low latency' for the FCC was something like
> > 100ms, and Musk was predicting <40ms.
>
> Indeed, the Rural Digital Opportunity Fund Auction defined "Low Latency"
> as "<100 ms", and "High Latency" as "≤ 750 ms & MOS of ≥4".

Except that it didn't define it under "normal working conditions" - as
in having any load on
that network, nor did it recognize that above 40Mbit, the bufferbloat
generally moves to the
wifi.

(not just to pick on starlink, most wifi setups are pretty broken, as
are most isps,
 as soon as you put a load on the network, even a single upload flow
messes up voip and videoconferencing)

so my hope is to move the FCC bar up in the coming months (I was just invited
to join BITAG). I do hope that we see better routers being mandated as a result
in the next funding round.

> MOS is the Mean Opinion Score, which takes latency and jitter into
> account. For "Low Latency", jitter/MOS was not defined. MOS goes from 1
> (Bad) to 5 (Excellent).

MOS is one of my favorite metrics, along with page load time.

The attached graph shows what we now achieve with airtime fairness and
the fq_codel scheduler and AQM
algorithms, in 5 shipping wifi chipsets, under load, and with multiple
stations present. From a MOS of 1, to over 4.
From the paper:

https://www.usenix.org/system/files/conference/atc17/atc17-hoiland-jorgensen.pdf

>
> https://www.fcc.gov/auction/904/factsheet
>
> FYI
> Daniel
> _______________________________________________
> Starlink mailing list
> Starlink@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink



-- 
Latest Podcast:
https://www.linkedin.com/feed/update/urn:li:activity:6791014284936785920/

Dave Täht CTO, TekLibre, LLC