10g residential CPE

Sat Dec 26 19:10:44 UTC 2020

I wouldn't rely on these numbers too much, your testing methodology is flawed.
People don't expect RING nodes to be used as speedtest servers and so they are usually not connected to high speed networks. 

Using a classical speedtest.net (Web or CLI) application would make much more sense, given the servers are actually connected to high speed Internet and are tuned to achieve such speeds - which is much more akin to how the most bandwidth demanding stuff (streaming, game downloads, system updates from CDNs) behaves. 

It's certainly possible to get 1G+ over >10ms RTT connections single stream - the buffers are certainly not THAT small for it to be a problem - not to mention game distribution platforms do usually open multiple connections to maximise the bandwidth utilisation. 

Re 85KB: that's just the initial window size, which will grow given tcp window scaling is enabled (default on modern Linux). 

Filip

On 26 December 2020 19:14:13 CET, Baldur Norddahl <baldur.norddahl at gmail.com> wrote:
>lør. 26. dec. 2020 18.55 skrev Mikael Abrahamsson <swmike at swm.pp.se>:
>
>> On Sat, 26 Dec 2020, Baldur Norddahl wrote:
>>
>> > It is true there have been TCP improvements but you can very easily
>> verify
>> > for yourself that it is very hard to get anywhere near 1 Gbps of
>actual
>> > transfer speed to destinations just 10 ms away. Try the nlnog ring
>> network
>> > like this:
>> >
>> > gigabit at gigabit01:~$ iperf -c netnod01.ring.nlnog.net
>> > ------------------------------------------------------------
>> > Client connecting to netnod01.ring.nlnog.net, TCP port 5001
>> > TCP window size: 85.0 KByte (default)
>> > ------------------------------------------------------------
>> > [  3] local 185.24.168.23 port 50632 connected with 185.42.136.5
>port
>> 5001
>> > [ ID] Interval       Transfer     Bandwidth
>> > [  3]  0.0-10.0 sec   452 MBytes   379 Mbits/sec
>>
>> Why would you just use 85KB of TCP window size?
>>
>> That's not the problem of buffering (or lack thereof) along the path,
>that
>> just not enough TCP window size for long-RTT high speed transfers.
>>
>
>That is just the starting window size. Also it is the default and I am
>not
>going to tune the connection because no such tuning will occur when you
>do
>your next far away download and wonder why it is so slow.
>
>If you do the math you will realise that 379 Mbps at 10 ms is
>impossible
>with 85 K window.
>
>I demonstrated that it is about buffers by showing the same download
>from a
>server that paces the traffic indeed gets the full 930 Mbps with
>exactly
>the same settings, including starting window size, and the same path
>(Copenhagen to Stockholm).
>
>Regards
>
>Baldur

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20201226/a4f97a62/attachment.html>