I'll have to connect the other router and run the test at some later date to get results. Speed wise, is about the same. There is some slowdown when running 2 consecutive speed test from different devises (mobile/desktop) But I don't know where to look to find the bottleneck. could be the switch, could be the modem, could be the rpi4. So long as we're no longer crashing the router during heavy(er) use.
I'll have to do more reading/research on how to setup SQM QoS properly.
nic = "nice" or "low priority" user process... it's probably at 0 (the number is to the left of the label) while idle is probably at 99%.
is that true even during 100Mbps download? I'd expect idle to drop to something, maybe 75% or like that during the peak download phase. Perhaps you can re-look now that you oriented to the fact that the number is to the left of the name
Wow, that's a significant amount of processor power if routing and firewall of multiple TCP streams at 100Mbps total costs only 3 or 4% of CPU!
Since I'm guessing you have DOCSIS 3.0 / 3.1 with the built-in queue management and therefore get reasonable bufferbloat stats without any SQM you might be all set as is. I do imagine SQM will increase the CPU usage a bit, but I'd still expect it to be 90% or 85% idle or something (part of it is there are 4 cores, and only one gets involved in queueing at any given time).
Enjoy your setup, I suspect we need someone with a 300+ Mbps stream and need for SQM before we will get numbers that help us estimate the performance envelope of the little Pi 4
Unfortunately, I don't know what DOCSIS is (yet). Running pretty much the stock image with only 1 mac address being restricted at the firewall and only LUCI and SQM being installed. I'll have to spend a few evenings setting things up so I can boot the kids out in the evenings.
DOCSIS is the cable modem standard (Data over Coaxial something something something). They added a queue manager system called "PIE" to the modems themselves.
You do have SQM going in this install already? If those numbers are for 100Mbps with cake I'm impressed. 3% of CPU is of course 12% of one core since one full core saturated would be 75% idle... still 12% of one core is not much and even if you used only one core, it can clearly do a bunch more packet processing.
Give them their own VLAN with their own SSID connected to that VLAN. Then you can easily have a firewall rule that prohibits forwarding that VLAN to the internet during certain time periods.
Because the issue seems to be the bridging table in the modem, which by default will only bridge one internal MAC to the docsis segment and is not persistent and hence will be relearned after a modem reboot, see details above.
