[Cake] CAKE tested on 3g connection in modem mode.

[Cake] CAKE tested on 3g connection in modem mode.

[Loganaden Velvindron]

Hi folks,

My setup:
tp-link archer c7 v2 with Huawei E355-s2 in modem mode, not Hilink
mode. I've use the AT commands to make modem mode permanent.

ISP: MTML Mauritius.

modem mode uses 3 as txqueuelen.

Before using the hilink mode:
https://www.dslreports.com/speedtest/4808641
https://www.dslreports.com/speedtest/4825996
https://www.dslreports.com/speedtest/4808742

Currently, I'm getting around 3.0Mbit/s, so I set it to 2.5Mbit/s, and
upload cap at 512Kbit/s, with an RTT of 500ms using "piece of cake".

Results (Bufferbloat score varies from A to B, and rarely A+. Due to
latency spikes, this is to be expected somewhat. I'm doing the test
using a laptop which uses fq_codel and txqueuelen of 0, which have
become the default for me since a while):

https://www.dslreports.com/speedtest/4832927
https://www.dslreports.com/speedtest/4832896

Re: [Cake] CAKE tested on 3g connection in modem mode.

[Loganaden Velvindron]

On Sun, Aug 28, 2016 at 8:23 AM, Loganaden Velvindron
<loganaden@gmail.com> wrote:
> Hi folks,
>
> My setup:
> tp-link archer c7 v2 with Huawei E355-s2 in modem mode, not Hilink
> mode. I've use the AT commands to make modem mode permanent.
>
> ISP: MTML Mauritius.
>
> modem mode uses 3 as txqueuelen.
>
> Before using the hilink mode:
> https://www.dslreports.com/speedtest/4808641
> https://www.dslreports.com/speedtest/4825996
> https://www.dslreports.com/speedtest/4808742
>
> Currently, I'm getting around 3.0Mbit/s, so I set it to 2.5Mbit/s, and
> upload cap at 512Kbit/s, with an RTT of 500ms using "piece of cake".
>
> Results (Bufferbloat score varies from A to B, and rarely A+. Due to
> latency spikes, this is to be expected somewhat. I'm doing the test
> using a laptop which uses fq_codel and txqueuelen of 0, which have
> become the default for me since a while):
>
> https://www.dslreports.com/speedtest/4832927
> https://www.dslreports.com/speedtest/4832896

It would be nice if the local ISPs could guarantee a minimum rate for
upload and download. That way, less hassle: I would just set it a bit
less than those values for sqm-qos, and I would be done with it.

In any case, I moved from 5000ms latency to approximately 500ms under
load, and honestly, I'm fairly happy for now. The mobile providers
could engineer their infrastructure so that there would be less
latencies variation.