Hey! Broadband peak time performance degradation

Hey all,

Those of you on HB, how's your evening peak time performance looking? I've started seeing significant degradation of performance, 8pm to 10pm every evening. 19th January evening for example was down to about 12Mb/s downstream which was terrible.

Performance 15th Jan to 21st Jan

Anyone else seeing this?

What are you using to monitor it?

It would be interesting to test whether the speeds are slow on the fibre itself, or out onto the internet only.

I use SpeedTest Tracker to monitor throughput and Uptime Kuma to monitor latency.

Both dead easy to spin up if you know Docker / Docker Compose

You can choose targets inside and outside of HBs Network.

Edited by nofappingway (Wed 22-Jan-25 13:49:11)

Well that just sent me down a rabbit hole!

Ok, I've managed figure out how Docker works, and I'm kinda [censored] that the Microsoft store has Ubuntu but not CentOS which I'm used to, but hey ho.

Set up the Speedtest Tracker, but I don't see any Hey!/F&W targets. I can see Zen, and my results are looking pretty abysmal already.

How can I set it up to look at Hey! and is there a way to alter the schedule? I found a video on Youtube, but I seem to be missing "Settings" on the left pane.

By default Speedtest Tracker chooses the 'fastest server' by the one that responds the quickest.

That video will be for an older version. The author moved config like that into the Docker Compose file (or Docker Run command if you used that) If you want to hardcode which server is used, you have to do that in your Docker Compose file or Docker Run cmd. The scanning schedule is configured there too. All the environmental options are here and example Docker Compose files are here

HB's server ID is 54208

For example, here's my Docker Compose file. Note that this forum will have removed all the indents that make this a valid YAML file:

services:
speedtest-tracker:
container_name: speedtest-tracker
ports:
- 8765:80
- 8443:443
environment:
- PUID=1001
- PGID=1001
- APP_KEY=base64:<your key here>
- DB_CONNECTION=sqlite
- SPEEDTEST_SCHEDULE=8 * * * *
- SPEEDTEST_SERVERS=54208
- PRUNE_RESULTS_OLDER_THAN=120days
- CHART_DATETIME_FORMAT=
- DATETIME_FORMAT=
- DISPLAY_TIMEZONE=Europe/London
volumes:
- ./config:/config
- ./config/keys:/config/keys
image: lscr.io/linuxserver/speedtest-tracker:latest
restart: unless-stopped

Edited by nofappingway (Wed 22-Jan-25 17:00:40)

Thanks, I've got it set up.

Very bizarre results though, have a look here: https://imgur.com/a/WSb0hXH

Completely different to running the Ookla app locally, which shows me getting ~900/900 to Hey!

Something isn't quite right there - all your downstream tests are virtually the same across different servers.

Agreed! Here's a web-based test to Zen run just now: https://imgur.com/gHHWIXj

So I guess the question is, if you run a manual test using the Ookla speedtest app, does it reflect the slow speeds you're seeing on the Speedtest Tracker at the relevant times.

Edited by AndyPandy (Wed 22-Jan-25 17:22:07)

Yeah, my Speedtest Tracker results match the Desktop Ookla client speedtests. I have Speedtest Tracker ping me a push notification when the result drops below a threshold I've set. At the same time, on Apple TV things buffer. online gaming lags etc. I. I've raised it with support.

I don't have a static IP so when I reboot my Router from time to time, I get a new public IP. I've observed the IP ranges differ drastically as does the performance for each one. This leads me to believe that there is some kind of load balancing going on as the first thing Support did was to forcefully bump me onto a specific network segment. That's performed well today so we will see what tonight brings.

In reply to a post by AndyPandy:

Ok, I've managed figure out how Docker works, and I'm kinda [censored] that the Microsoft store has Ubuntu but not CentOS which I'm used to, but hey ho.

CentOS is end of life, version 7 is out of support, and version 8 was ended. CentOS Stream is not the same thing. Rocky Linux is the closest to what CentOS was as a free version of RHEL, but most people don't want server linux. Desktop Linux is better with Ubuntu or even Fedora.

You can run Rocky 8 or 9 in a Hyper-V VM easily enough if you want a free "very close to RHEL" Linux.

What does your BQM look like? I would expect any congestion to become apparent from the BQM.

I'm not using it. I'm using my own outward looking tools mentioned above to monitor throughput, latency and packet loss.

Here's my tracker: https://imgur.com/p0uFIp9

I'm not seeing the dips that you are, although my download speeds within docker are about 1/4 of what I get using the Ookla app in Windows.

Thanks for following up Andy. HB Support moved me to a different network and since then it's been all good.

Hats off to their support, they acted quickly and didn't muck me around.

In reply to a post by nofappingway:

Thanks for following up Andy. HB Support moved me to a different network and since then it's been all good.

Hats off to their support, they acted quickly and didn't muck me around.

Presumably you mean logically, in their upstream network, rather than at the PON-level?

Correct, different subnet. I can only assume they are load balancing here.