A.28 Monitor of QoS flow release

28.5523GPP5G performance measurementsManagement and orchestrationRelease 18TS

QoS flow is the key and limited resource for 5G RAN (including NG-RAN and non-3GPP access) to deliver services. The release of the QoS flow needs to be monitored as:

– an abnormal release of the QoS flow will cause the call(/session) drop, which directly impacts the QoS delivered by the networks, and the satisfaction degree of the end user;

– a successfully released QoS flow can be used to setup other requested calls(/sessions). The QoS flow failed to be released will still occupy the limited resource and hence it can not be used to admit other requested calls(/sessions).

From a retainability measurement aspect, QoS flows do not need to be released because they are inactive, they can be kept to give fast access when new data arrives.

To define (from a QoS flow release measurement point of view) if a QoS flow is considered active or not, the QoS flow can be divided into two groups:

For QoS flows with bursty flow, a UE is said to be "in session" if there is user data in the PDCP queue in any of the directions or if any QoS flow data on a Data Radio Bearer (UL or DL) has been transferred during the last 100 ms.
For QoS flows with continuous flow, the QoS flow (and the UE) is seen as being "in session" in the context of this measurement as long as the UE is in RRC connected state, and the session time is increased from the first data transmission on the QoS flow until 100 ms after the last data transmission on the QoS flow.

A particular QoS flow is defined to be of type continuous flow if the 5QI is any of {1, 2, 65, 66}.

The specific reason causing the abnormal and failed release of the QoS flow is required in order to find out the problem and ascertain the solutions. And due to different priority and tolerance for different service type with different QoS level in the networks, the monitor needs to be opened on each service type with QoS level.

The QoS flow can be released by PDU Session Resource Release procedure, UE Context Release procedure, Reset procedure either initiated by 5G RAN (including NG-RAN and non-3GPP access) or AMF and NG Path Switch procedure (see TS 38.413 [11]).

So performance measurements related to QoS flow Release (see TS 38.413 [11]) and UE Context Release (see TS 38.413 [11]) procedure for each service type with QoS level are necessary to support the monitor of QoS flow release.

The abnormal release of the QoS flow has potential scenario where, regardless of receiving the UE Context Release Command with the cause related to abnormal release, the end user does not perceive it as abnormal. This scenario is explicitly related to 5QI 1 calls, for other services it is not possible to determine the reason behind the cause code. It is typical to encounter such scenario, a so called "double UE Context", when Radio Link Failure occurs during an ongoing 5QI 1 call and RRC Connection Re-establishment attempt fails on target or other cell. If then the UE does a new RRC Connection the 5QI 1 QoS flow is set-up during Initial Context Setup in the target or other cell. However, when AMF receives that service request with the Initial UE message through the target or other cell, it realizes that it already has the same UE Context but from the source cell (it has not been released yet). In such case, AMF sends UE Context Release Command to the source cell. As the 5QI 1 QoS flow has been successfully setup in the target or other cell, the 5QI 1 QoS flow release in the source cell may not be perceived as a drop (abnormal release) by the end user, as the service has been sustained with some interruption time, and can’t be considered as a drop in the 5QI 1 QoS flow Drop Ratio. This interruption time may be monitored in order to evaluate how it can impact the QoS of the 5QI 1 Flows due to double NG (double UE context)". Moreover, the 5QI 1 QoS Flows that can be immediately released due to radio reasons with UE connectivity lost (when T-RLF timer was not started) may be delayed by time interval based on this average interruption time to possibly transform them to double NG scenario to keep the calls active and reduce further the 5QI 1 QoS flow Drop Ratio.

From QoS perspective it is important to focus also on call duration as in some cases wrong quality perceived by the end user is not fully reflected by drop ratio nor retainability KPI. Typical case is when due to poor radio conditions the end user redials (the call was terminated normally) to the same party to secure the quality. But in this case the drop ratio KPI will not show any degradation. Secondly, although the call is dropped the end user may or may not redial depending on dropped call duration compared to the case when the call would be normally released. It is therefore highly recommended to monitor average and distribution of duration of normally and abnormally released calls.