Here we go again! My ISP had the 2nd network failure in a week 😬

May i ask where you got this statistics graph?

@jonas-w I installed Grafana on a raspberry. The plugin name is "Raspberry Pi Monitoring" by Jorge de la Cruz

I also want to know how you gathered this data. I had to make my own speed testing application when I once wanted to monitor my home internet speed for a period of time.

@davide Oh, ok there's my answer ;)

Well for a right speed test use speedtest-cli or something like that. You measure the values and then you write in a influxdb database. Check also Network Monitoring on the grafana plugin website

@davide i would use librespeed. I heard nothing good about speedtest by ookla

Y u up so late

It’s dead Jim

@jonas-w can you elaborate?

@davide i guess it's not directly speedtest.net. But how some ISPs treat speedtest.net by allowing full bandwith to speedtest.net and "faking" the results. I don't know if that is still the case. Just heard that somewhere and it's the internet and not all stuff should be trusted blindly. Never had that experience first hand. But if i do speedtests i always use librespeed mostly because of privacy.

Okay well i just tried speedtest.net and had a ping of 0???

@jonas-w mh interesting, never heard such a thing but I will investigate. In any case my ping is 12 ms

@davide okay i just tried the cli version and it seems like the web version is a bit weird.

CLI: 30-35ms ping, 10-12 download and 4-6 upload

Web: always 0ms ping, 20-25 download and 10-15 upload

i tried several times and always got these results.

It also selects every time the same server so that doesn't seem to be the culprit

@jonas-w Librespeed seems pretty good, although their test is fairly short, and seems sometimes heavily bottlenecked by the choice of the remote.

On paper, I should currently get 1Gbit, and usually get 700-900 Mbit/s in practice.

Librespeed reports in the 600-700 range to a server in Amsterdam (close to me), but only 30-40 to Frankfurt.

I regularly download things from Frankfurt (AWS/Gcloud servers), and get rates up to 900 just fine in that case.

@bittersweet probably because those vps themselve dont have 1gbit, it would probably be good to write their theoretical speed next to the servers.

@jonas-w Yeah. Reliable speedtesting is actually a pretty hard problem to solve, especially for free -- I mean you're basically inviting other people to saturate your bandwidth with a random burst of data.

But the effort to move away from commercial tests is very commendable, and it works well enough 🤷‍♀️