SQM QOS - Slow speeds

Hello,

I don't have a fast connection to begin with. I have a WISP that provides me with 50Mbit/s down and 10Mbit/s up. I've intentionally set my ingress to 31200 and my egress to 8000, as sometimes the available bandwidth can drop down during peak usage, and I want to compensate for that. With SQM off, I average somewhere around 42Mbit/s down and 9.25 Mbit/s up. With SQM on, I average somewhere around 21.3 Mbit/s down and 7 Mbit/s up. The router is a Pi4 running on 21.02. It doesn't even break a sweat when running the speedtest. Direct ethernet connection to the antenna, no modem. I've tried playing with the overhead. I set it to 22 and also set it to none. It didn't seem to make any difference. Any help would be appreciated.

top during speedtest with SQM on
image

config
image

1 Like

SQM is actually for latency, but my guess if you are saturating your upload with ACK's since you are on an asymmetric line.

Try setting your download to 0 disabling it,
your upload to 10000,
and memlimit 32MB ack-filter wash for your advanced options for egress.

Start here and then report back.

SQM ON
ingress 0
egress 10000
memlimit 32MB ack-filter wash

SQM On
ingress 0
egress 10000

SQM Off

Thanks for testing. Those numbers are all over the place. I'm not real familiar with WISP connections but you would sharing with others also connected to the same point, similar to an AP or cell phone tower.

You mentioned you get 42Mbps down and 9.25 Mbps up with no SQM but the test showed 25.8 / 9.53. Are you getting different results at different times of the day?

Yeah, it get's worse in the evenings during peak hours, and it's inconsistent. That's why I had set my ingress to 31k, but it does go below that at times. Of course my ISP says all of their remote tests "look good".

For truly variable rate links you might want to try:

Thanks, I will give that a shot!

I got the script set up, and so far it's working great! I set my min to 15000, baseline to 43000 and max to 50000. Delay is set to 40. I'm consistently getting an A rating when testing for bufferbloat on waveform. Appreciate all the work that went into writing this script and for the help! I do have one IP that I'm routing via PBR to an interface that's a wireguard client that connects to PIA. Do you see any value in trying the cake-dual-ifb script?

I would refer this question to @Lynx, as he wrote and used that script. I could only waffle on, on theoretic grounds.
For questions regarding autorate, you might want to bring them up in the main autorate thread, that is if you have any....

So I think the situation where cake-dual-ifb or cake-wg-pbr might be helpful is where your really want cake to see within what is otherwise encrypted. Say you have five LAN flows and four go over the VPN then if cake can only see the two flows (one unencrypted and one encrypted) then it can't provide flow fairness between the five flows. Does that situation apply to you? If so, then either of the above scripts may he useful to you and worth setting up.