Some data from my automated monitoring. Speed is measured with an Ookla speed test every 15 minutes or (now) every hour: daily min, max, and average. Latency is measured with telegraf native pings to 8.8.8.8: daily min, average. Packet loss is the daily average percentage of ping packets lost. There’s some messy data here; sometimes my house fails over to a 12Mbps fixed wireless link which pollutes these measurements (but only a little). And there’s a suspiciously low bandwidth stretch in December and January that could well be a monitoring bug.
Important to read these right. These are daily average statistics: the min/avg/max bandwidth for each day, the min/avg ping time for each day. Each latency measurement is actually several pings sent at once, so I’m taking the average of the average and the min of the min.
The primary pattern that jumps out here is the increase in latency starting in January. That coincides with when I got moved from a Northern California POP up to the Seattle POP (I think via a satellite relay near Winters, CA). Although note the change is gradual; this isn’t a one-off bump in path length. Another possible explanation is that’s the period Starlink started adding a lot more users and I’m seeing buffering.
The min bandwidth graph does not bear out my perception, which is that congestion is getting worse. It’s definitely getting slower every night. Below is a graph of average download speed for 15 minute samples over a week in January; the pattern now is about the same. But the minimum daily I’m getting is still in the 10-20Mbps range just like it’s always been.
Nelson's log 