Thanks for sharing this sad news here, Adam. Very much appreciated.
In the living memory of Dave today I decided to tune my SQM to the edge of possibility using cake ~5ms latency increase under full load. I sacrificed some part of the full speed but stability is noticably improved now.
Schnapps snapshot made just in case.
I hacked a bit betterspeedtest.sh from Openwrt scripts from github.
root@router:~# bufferbloat.sh
2025-04-19 16:32:06 Testing against netperf-eu.bufferbloat.net (ipv4) with 16 simultaneous sessions while pinging 1.1.1.1 (30 seconds in each direction)
.............................
Download: 326.64 Mbps
Latency: (in msec, 31 pings, 0.00% packet loss)
Min: 20.200
10pct: 20.300
Median: 20.500
Avg: 20.819
90pct: 20.900
Max: 25.600
..............................
Upload: 192.23 Mbps
Latency: (in msec, 31 pings, 0.00% packet loss)
Min: 20.200
10pct: 20.200
Median: 20.400
Avg: 20.413
90pct: 20.500
Max: 20.900
Could I ask you to not share this hack, please? Rich Brown pays for the VPS at the other end out of his own pocket, and he introduced that password to slow down automated tests, so that the included data volume does not only last a few days at the beginning of the month…
I am confident that your usage to tune SQM is what Rich wants to enable, just not regular scripts to measure throughput.
Sure thing! I just used that to tune cake. You have my word but it was way to easy to do.
Actually @moeller0 many of your posts on OpenWRT forum were helpful.
Edit:
I asked Rich via e-mail about donation options to stay fair🥳