How Maximum Bitrate (MBR) via QER is applied?

Q/A
#1

Dears, I was investigating on how the QoS profile is effectively applied by the UPF. I looked at the wireshark dump on N4, in which I can see the MBR/GBR rules arriving, as well as the source code, in which dialogue with N4 and with the GTP kernel driver takes place and is understood. After QER MBR is applied to the UP, running iperf flows between UERANSIM’s UE and DN result into an effective bandwidth limitation which works.

The point is that I cannot find any point neither in UPF nor in the GTP kernel module where traffic shaping is enforced. Could you please redirect me to the correct place?

Secondly, it appears that if I update QER rules via libgtp5gnl commands (e.g., increaseing MBR value), change is not effective and previous bandwidth limit is still in place.
(rel 3.0.6).

Thanks!

F.

#2

Hi @francozamp,

Currently, the implementation of the gtp5g kernel module will receive the QER information from UPF, and also at most 4QER only per Sesssion. We will use the QFI information from QER and fill it in downlink packet which is towards to gNB.

Right now, there is no plan to implement QoS in the gtp5g kernel module, and also looking for support from the community to implement QoS in the kernel module.

#3