Woolf

Hello Keenetic team and community, I understand that KeeneticOS uses CAKE-based AQM/SQM, and that NFQ is used for shaping at higher speeds while CAKE is used at lower rates. What I would like to understand better is which CAKE/SQM-related mechanisms are actually active internally, and how they are implemented. Could someone from the Keenetic team or experienced users confirm whether the following mechanisms are active in the current IntelliQoS/SQM/NFQ implementation? A simple yes/no or short explanation for each item would be very helpful. NAT-aware host fairness Can SQM fairly separate LAN devices behind NAT? ACK filtering Is TCP ACK filtering, or any similar ACK optimization, used? WAN overhead compensation Does SQM account for PPPoE, VLAN, DSL, cable or Ethernet overhead while shaping? Download-side SQM / IFB Is download shaping done with an ingress / IFB-like mechanism, or with another Keenetic-specific method? Packet marking / priority mapping After IntelliQoS classifies traffic, are packets marked with DSCP/802.1p, or are they only mapped internally to priority queues? How are CAKE features limited or simplified in NFQ to ensure fast loading speeds? Does IntelliQoS / NFQ protect small latency-sensitive UDP flows under heavy load independently of application recognition, at the SQM/queueing level or through Keenetic’s own scheduler? My goal is not to criticize IntelliQoS. I am trying to understand how close the current implementation is to typical OpenWrt SQM-CAKE behavior in areas such as host fairness, overhead compensation, ACK filtering and download-side shaping. Any explanation from moderators, developers or experienced users would be appreciated. Thank you.

Hello Keenetic team, I would like to request a more transparent and more flexible SQM option inside IntelliQoS. I understand that IntelliQoS already provides application classification, traffic prioritization and CAKE-based SQM/AQM behavior. However, some users may prefer an OpenWrt like CAKE-only SQM mode that focuses mainly on queue control, fairness and bufferbloat reduction, without depending on application classification or an NFQ/classification-based traffic path. My request is not to remove IntelliQoS. Instead, please consider adding an optional mode such as: IntelliQoS automatic mode; IntelliQoS with application prioritization; CAKE-only SQM mode. The CAKE-only mode would be useful for users who want predictable latency control even when traffic is encrypted, unknown, mixed, VPN-based, or not classified correctly. In many real home networks, perfect application recognition is difficult. It would also be useful to document or expose which CAKE/SQM mechanisms are active, such as: device/host fairness; flow fairness; DiffServ behavior, including modes similar to besteffort, diffserv3 or diffserv4; NAT awareness; ACK filtering; WAN overhead compensation for PPPoE, VLAN, DSL, cable and Ethernet; download-side SQM using ingress shaping / IFB-like behavior; whether SQM is active on upload, download or both. The main practical benefit would be a choice between two approaches: IntelliQoS mode: application-aware prioritization and automatic classification. CAKE-only SQM mode: advanced, more deterministic queue management focused on fairness and stable loaded latency. This would make KeeneticOS more attractive to advanced users who currently prefer OpenWrt SQM-CAKE for predictable bufferbloat control, host fairness and transparent CAKE behavior. Thank you for considering this request.