SQM, QoS, CAKE and fq_codel: what to try
Smart Queue Management can reduce bufferbloat by controlling the bottleneck and keeping queues short. Here is a safe, measurement-first way to test it.
SQM vs traditional QoS
Traditional QoS often prioritizes selected traffic. SQM focuses on queue management and typically combines shaping with modern queue disciplines such as CAKE or fq_codel. The goal is not to make the line faster; it is to stop bulk traffic from creating large delays for interactive traffic.
A practical tuning loop
- Run Connection Vitals and save a Baseline.
- Measure your real download and upload rates over a stable wired connection if possible.
- Enable the router's SQM / Smart Queue / CAKE / fq_codel feature if supported.
- Start shaping around 90–95% of the measured line rate.
- Run Connection Vitals again with the label “After SQM/QoS”.
- Compare loaded latency and throughput. Increase the limits carefully if responsiveness stays good and you want to recover some throughput.
Why 90–95%?
The router needs to become the place where the queue is controlled. If it shapes above the real bottleneck rate, queues may still build elsewhere. The ideal setting depends on the connection and can vary, so 90–95% is a starting experiment—not a universal optimum.
If your router has no SQM
First confirm that bufferbloat is actually the problem. Firmware, ISP hardware, bridge mode and router replacement can all be options, but changing hardware should not be the first step when a configuration change might solve it.