[Cake] [Rpm] goresponsiveness over slow 5g

[Will Hawkins]

Thanks for the report. We are working on updates once we can get some
upstream issues resolved with Go. Sorry for the slow progress -- doing
the best that we can!

Will

On Wed, Apr 6, 2022 at 12:17 PM Dave Taht via Rpm
<rpm at lists.bufferbloat.net> wrote:
>
> Pulled down the latest
> https://github.com/network-quality/goresponsiveness and built it with
> go.1.18.
>
> Some notes in #s. Tests were performed via linux kernel
>
> Linux milliways 5.4.0-80-lowlatency #90-Ubuntu SMP PREEMPT Fri Jul 9
> 23:56:54 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
>
> over tethered wifi to a oneplus nord n105g running oxygen OS
> 11.05BE.86AA, linux kernel 4.19.125perf+
>
> # No cake on the wifi:
>
> ./goresponsiveness --config mensura.cdn-apple.com --port 443 --path
> /api/v1/gm/config
> 04-06-2022 15:52:32 UTC Go Responsiveness to mensura.cdn-apple.com:443...
> Download:   7.594 Mbps (  0.949 MBps), using 28 parallel connections.
>
> # Really should cap this. 28 flows at this rate is certain to end in tears.
>
> Upload:     1.906 Mbps (  0.238 MBps), using 12 parallel connections.
> Total RTTs measured: 5
> RPM:   170
>
>
> # cake 6mbit/3mbit
>
> davet at milliways:~/git/goresponsiveness$ ./goresponsiveness --config
> mensura.cdn-apple.com --port 443 --path /api/v1/gm/config
> 04-06-2022 15:53:57 UTC Go Responsiveness to mensura.cdn-apple.com:443...
> Download:   8.343 Mbps (  1.043 MBps), using 12 parallel connections.
> Upload:     2.500 Mbps (  0.312 MBps), using 12 parallel connections.
> Total RTTs measured: 5
> RPM:   182
>
> ./goresponsiveness --config mensura.cdn-apple.com --port 443 --path
> /api/v1/gm/config
> 04-06-2022 15:55:11 UTC Go Responsiveness to mensura.cdn-apple.com:443...
> Download:   6.750 Mbps (  0.844 MBps), using 12 parallel connections.
> Upload:     2.000 Mbps (  0.250 MBps), using 36 parallel connections.
>
> # 36??? flows at this rate is utterly certain to end in tears.
>
> Total RTTs measured: 5
> RPM:   154
>
> # cake output after these testss
>
>                   Tin 0
>   thresh          3Mbit
>   target          6.1ms
>   interval      101.1ms
>   pk_delay       14.7ms
>   av_delay        6.4ms
>   sp_delay          8us
>   backlog            0b
>   pkts            15275
>   bytes        10576809
>   way_inds          252
>   way_miss          191
>   way_cols            0
>   drops            2807
>   marks               0
>   ack_drop            0
>   sp_flows           14
>   bk_flows            1
>   un_flows            0
>   max_len          6470
>   quantum           300
>
> # RFC3168 style ECN IS enabled on the mensura server. Yay!
> # If I enable it on my client
>
> ./goresponsiveness --config mensura.cdn-apple.com --port 443 --path
> /api/v1/gm/config
> 04-06-2022 16:03:47 UTC Go Responsiveness to mensura.cdn-apple.com:443...
> Error: Saturation could not be completed in time and no provisional
> rates could be accessed. Test failed.
>
>                  Tin 0
>   thresh          6Mbit
>   target          5.0ms
>   interval      100.0ms
>   pk_delay       23.3ms
>   av_delay       13.6ms
>   sp_delay         16us
>   backlog            0b
>   pkts            40856
>   bytes        34440011
>   way_inds          774
>   way_miss          402
>   way_cols            0
>   drops             956
>   marks            8907
> # mark away!
>   ack_drop            0
>   sp_flows           63
>   bk_flows            1
>   un_flows            0
>   max_len          1442
>   quantum           300
>
> # with ecn, repeated tests almost always timeout. It would be useful
> if they displayed their attempted params...
>
> ./goresponsiveness --config mensura.cdn-apple.com --port 443 --path
> /api/v1/gm/config
> 04-06-2022 16:10:30 UTC Go Responsiveness to mensura.cdn-apple.com:443...
> Download:   4.562 Mbps (  0.570 MBps), using 12 parallel connections.
> Upload:     2.594 Mbps (  0.324 MBps), using 44 parallel connections.
>
> # this inversion between lower bandwdith and higher connects needs to
> get looked at.
>
> Total RTTs measured: 5
>
> # perhaps a goal of trying to measure for more RTTs than five would help.
>
> --
> I tried to build a better future, a few times:
> https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org
>
> Dave Täht CEO, TekLibre, LLC
> _______________________________________________
> Rpm mailing list
> Rpm at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/rpm