Dynamic quality of service prioritizing in mobile networks

Abstract

This disclosure relates a dynamic quality of service (QoS) prioritization in a mobile network. A network entity is provided, which receives input information comprising a number of priority levels and / or a network congestion level, and provides, in a current time period, output information comprising a first QoS priority setting for a first future time period. The first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow, or for each of two or more QoS flows of the service group. A consumer network entity is further provided, which receives the output information of the network entity, and applies, in the first future time period, the priority levels to the two or more packet types of the QoS flow, or to the two or more QoS flows of the service group.

Description

[0001] DYNAMIC QUALITY OF SERVICE PRIORITIZING IN MOBILE NETWORKS

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to mobile communication networks, for instance, to 5thgeneration (5G) or 6thgeneration (6G) mobile networks. The disclosure is concerned with quality of service (QoS) of services provided in the mobile network. The disclosure is especially concerned with supporting a dynamic QoS prioritization. To this end, the disclosure present a network entity for determining a prioritization of packet types of a QoS flow, and / or for determining a prioritization of QoS flows of a service group with multiple QoS flows. The disclosure also presents a consumer network entity for applying the determined prioritization, and presents corresponding methods.

[0004] BACKGROUND

[0005] Various new video-based media services are emerging with the advance of tele-communication technologies: e.g., cloud virtual reality (VR) or augmented reality (AR), cloud gaming, 4K / 8K video on demand, video-based remote control, and machine vision. Such services require not only a high bandwidth, but also a short end-to-end (e2e) latency and high communication reliability. For instance, the consumption of VR content via a tethered VR headset requires a maximum 5-10 ms e2e latency, a 0.1-10 Gbit / s service bit rate, and 99.99% reliability (see, e.g., TS 22.261vl8.2.0). Atransfer of such type of service traffic is challenging, especially in mobile networks, due to the limited radio resources and the dynamic radio conditions.

[0006] To reduce the required bandwidth for transferring video streams, various compression technologies have been studied. For instance, H.264 compresses a video stream by exploring the temporal / spatial dependency of the video traffic. H.264 introduces the concept of group of picture (GOP) with different types of video frames. Intra-coded (I) frames contain entire images. I-frames are coded without reference to any other frame. Predicted (P) frames reference to proceeding frames for decoding and / or prediction. Bi-directional predicted (B) frames reference to proceeding and subsequent frames to be displayed. The dependency relationship decides the importance of the frames in the video stream for the application, e.g., I > P > B in this case, and accordingly the importance of packets corresponding to the frames. Similarly, packets in a base layer are more important than the packets in an enhanced layer in case of layered coding (as shown in FIG. 1). As another example, in the case of 360° videos, the scene at the central vision area of the field of view (FOV) is more important than the scene at the edge of the FOV.

[0007] Mobile networks may use such importance information for a prioritized treatment of packets within XR (i.e. VR or AR) traffic. For example, mobile networks can schedule the transmission of the more important packets first. In case of a network congestion, the less important packets will thus have a higher chance to be dropped instead of the more important packets, which is due to the expiration of the packet delay budget (PDB) or of the protocol data unit (PDU) set delay budget (PSDB), respectively. In this way, the impact to a user observed quality of experience (QoE) can be minimized, and more XR services can be supported by a certain amount of radio resource in the mobile network with sufficient user satisfaction.

[0008] However, it is not always possible to get the importance information of the packets, for instance, due to application layer encryption.

[0009] Notably, in this disclosure a PDU set refers to one or more PDUs that together carry the payload of one unit of information (corresponding to a packet in the network layer), which is generated at the application level (e.g., an I-frame). A PDU set is also named media unit (MU) in this disclosure. That is, an MU may correspond to a packet. The MU type describes the type of the MU, for example, of an XR service (e.g., an 1-frame or P-frame). The MU type may correspond to a packet type.

[0010] Meanwhile, a prioritized treatment in the mobile network may come with additional overhead. For example, the detection of important packets would need the inspection of the packet headers of a service flow, which may cause an additional delay in the user plane transport in the mobile network. As another example, classification of service traffic into different processing queues may also introduce additional complexity of queue management.

[0011] Such types of overhead affect again the QoS / QoE of the user service, and becomes unnecessary in a good network condition (i.e., if the network is not congested). SUMMARY

[0012] The present disclosure and its solutions are based further on the following considerations.

[0013] 3GPP Rel. 17 defines that QoS of mobile communication is managed and controlled per QoS flow(s). As shown in FIG. 2, application / service layer packets are mapped into one or more QoS flows using a packet filter. The packet filter consists of 5 tuples (i.e., source / destination IP address, source / destination port, application ID). Each QoS flow is bound to a QoS profile, which is derived from the application layer service requirements. The network entities (e.g., radio access network (RAN) entity, session management function (SMF) entity) treat each QoS flow separately according to the corresponding QoS profile.

[0014] The QoS profile includes a priority level as one of the QoS characteristics. However, the current QoS model does not support prioritizing packets (e.g., PDUs of a PDU set) within QoS flows. QoS adaptation based on the contents of service traffic is also not possible.

[0015] 3GPP Rel.18 works on specifying the P-trace from an XR application server to a mobile network, which would include the importance information of the packets (e.g., PDUs of a PDU set), and will also be visible to the mobile network. However, it may not be appropriate to use such information directly in the mobile network for QoS prioritization considering the overhead of the related processing as explained before.

[0016] Overall, existing solutions are either not able to prioritize packet treatment based on packet importance from the application layer, or are inefficient due to a lack of consideration of dynamic network conditions (e.g., network load and / or congestion) and overhead of the related network treatment.

[0017] In view of the above, this disclosure aims to provide an improved solution. An objective is to enable efficient prioritization of packets of a QoS flow, or efficient prioritization of QoS flows of a service group, wherein the prioritization is based on importance from the application layer. At the same time, an objective is to consider dynamic network conditions and overhead when determining the prioritization. A specific objective is to detect the importance of the packets of a QoS flow, or the importance of QoS flows of a service group, without or with partial importance information from the application. Another objective is to determine the best prioritizing strategy in the mobile network, in order to maximize the overall service capacity, for example, to retain sufficient user satisfaction under different network conditions. Another objective is to determine the best QoS strategy jointly considering the network condition (network load and / or congestion) and related processing overhead in the network. Another objective it to adapt the QoS strategy based on the dynamic contents of the application.

[0018] These and other objectives are achieved by the solutions as described in the independent claims. Advantageous implementations are further defined in the dependent claims.

[0019] A first aspect of this disclosure provides a network entity for determining a prioritization of packet types of packets of a QoS flow, the network entity being configured to: receive input information comprising a number of priority levels and / or a network congestion level; and provide, in a current time period, output information comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow.

[0020] The network entity of the first aspect is able to determine the prioritization (the respective priority levels) of the packet types of the QoS flow based on the number of priority levels provided, and considering dynamic network congestion levels reflecting the network load and / or congestion. The determined prioritization may maximize the overall service capacity, and may retain sufficient user satisfaction under different network conditions. The network entity of the first aspect may consider both the network load and / or network congestion, and may consider a related processing overhead in a mobile network, when determining the prioritization. The network entity of the first aspect may also adapt the first QoS priority setting based on new input information, for instance, changed due to dynamic contents of the application related to the QoS flow. Notably, prioritizing packet types refers to determining the prioritizing strategy for these packet types, not to enforcing the prioritization. The enforcing may be done by another entity.

[0021] The “network” in “network load” and “network congestion” refers to the part of the network, in which the flow(s) of the services are accommodated.

[0022] A second aspect of this disclosure provides a network entity for determining a prioritization of QoS flows of a service group, the network entity being configured to: receive input information comprising a number of priority levels and / or a network congestion level; and provide, in a current time period, output information comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more QoS flows of the service group.

[0023] The network entity of the second aspect is able to determine the prioritization (the respective priority levels) of the QoS flows of the service group of QoS flows based on the number of priority levels provided, and considering dynamic network congestion levels reflecting the network load and / or congestion. The determined prioritization may maximize the overall service capacity, and may retain sufficient user satisfaction under different network conditions. The network entity of the second aspect may consider both the network load and / or network congestion, and may consider a related processing overhead in a mobile network, when determining the prioritization. The network entity of the second aspect may also adapt the first QoS priority setting based on new input information, for instance, changed due to dynamic contents of the application related to the service group. Notably, prioritizing QoS flows refers to determining the prioritizing strategy for these QoS flows, not to enforcing the prioritization. The enforcing may be done by another entity.

[0024] The network congestion level in this disclosure may be equivalent to a flow congestion level (e.g., in case of prioritizing a single QoS flow), and may be equivalent to a session congestion level (e.g., in case of prioritizing two or more QoS flows of the service group belonging to the same session).

[0025] In an implementation form of the first or second aspect, the network entity is further configured to determine the first QoS priority setting based on the received number of priority levels and / or the received network congestion level.

[0026] For instance, the network entity may comprise a processor configured to determine the output information based on the input information.

[0027] In an implementation form of the first or second aspect, the input information comprises a respective number of priority levels for each of two or more network congestion levels. In this way, the network entity of the first or second aspect is able to take into account the current network congestion and / or load for determining the first QoS priority setting. Depending on the network congestion level, the first QoS priority setting may indicate different priority levels for the packet types or the QoS flows.

[0028] In an implementation form of the first or second aspect, the network entity is further configured to: determine a network congestion level and / or a QoS fulfilment status for the first future time period; and determine the first QoS priority setting according to the determined network congestion level and / or the determined QoS fulfilment status.

[0029] Thus, the current QoS is taken into account. The QoS fulfilment status indicates to what extent the target QoS is currently realized in the mobile network. The QoS fulfilment status may indicate one of multiple QoS levels, or may indicate whether the target QoS is fulfilled or not, or may indicate whether a certain QoS parameter in the target QoS is fulfilled or not fulfilled.

[0030] In an implementation form of the first or second aspect, the network entity is configured to: receive a network congestion level and / or a QoS fulfilment status for the current time period by indication from another network entity; and determine the network congestion level and / or the QoS fulfilment status for the first future time period based on the network congestion level and / or QoS fulfilment status for the current time period.

[0031] For instance, the network entity may comprise a processor, which is configured to predict the network congestion level and / or the QoS fulfilment status for the first future time period.

[0032] In an implementation form of the first or second aspect, the input information further comprises at least one of: an application ID; a flow ID or flow description; a current quality of experience, QoE, of an application related to the QoS flow or to the service group; a second QoS priority setting for the current time period, wherein the second QoS priority setting indicates one of the priority levels for each of the two or more packet types of the QoS flow or for each of the two or more QoS flows of the service group; an indication of one or more packet types of the QoS flow: a service group ID of the service group; a QoS fulfilment status for the current time period, wherein the QoS fulfilment status is related to one or a combination of QoS parameters of a QoS flow selected from: a packet error rate, a packet delay budget, a protocol data unit set error rate, and / or a protocol data unit set delay budget, of the packets of the QoS flow or of the QoS flows of the service group. In this disclosure, an application ID may be used to indicate a service group. Further, a flow ID or flow description may be used to indicate a QoS flow.

[0033] In an implementation form of the first or second aspect, the network entity is further configured to receive application layer priority information indicating the two or more packet types.

[0034] In an implementation form of the first or second aspect, the network entity is configured to repeatedly update and provide the output information every time period or every time the output information changes.

[0035] In an implementation form of the first or second aspect, the network entity is configured to determine the output information using a trainable model.

[0036] In an implementation form of the first or second aspect, the network entity is configured to obtain the trainable model from another network entity and / or train the trainable model based on the input information.

[0037] That is, the model used by the network entity to determine the output information may be pretrained, and / or may be trained by the network entity itself based on machine learning techniques The model may comprise a neural network, for instance, a convolutional neural network (CNN) and / or deep learning neural network (DNN).

[0038] In an implementation form of the first or second aspect, the network entity is configured to: obtain a relationship between the first QoS priority setting for the first future time period and a QoE during the first future time period of an application related to the QoS flow or to the service group; and determine and provide the output information for a second future time period based on the relationship and based on an award to increase the QoE.

[0039] The learned relationship between the first QoS priority setting and the QoE may be updated by the network entity, for instance, periodically or occasionally. Based on the updated relationship, the output information may change. In this way, a recursive re-determination of the output information may be performed. Additional future time periods are thus envisioned, beyond the first and second future time period. In an implementation form of the first or second aspect, the trainable model comprises a plurality of sub-models, wherein each sub-model is configured for a different number of priority levels; and the network entity is configured to select one of the sub-models based on the number of priority levels included in the input information, and to use the selected sub-model for determining the output information.

[0040] For instance, each sub-model is related to a different upper bound of the number of priority levels. Upper bound means that, for instance, if the number of priority level for a particular submodel is 3, then the number of priority levels may be 2 or 3. However, the network entity may increase the priority level of one or more sub-models, for instance, stepwise and unbounded by the input information.

[0041] In an implementation form of the first or second aspect, the network entity is further configured to select the one of the sub-models based on the relationship and based on the award to increase the QoE.

[0042] In an implementation form of the first or second aspect, the input information further comprises a progressive priority level indication, and according to the progressive priority level indication, the network entity is configured to update at least one of the sub-models by stepwise increasing the number of priority levels related to the sub-model; or the network entity is configured to select the one of the sub-models based further on an increased number of priority levels.

[0043] In an implementation form of the first aspect, the network entity is configured to provide the output information to a consumer network entity, for example, a policy control function (PCF) entity or a SMF entity.

[0044] In an implementation form of the first aspect, the network entity is configured to: receive the number of priority levels or a respective number of priority levels for each of two or more network congestion levels included in the input information from an OAM entity; and / or receive at least one of the second QoS priority setting, the QoS fulfilment status, and the network congestion level included in the input information from a PCF entity, or a SMF entity, or a RAN entity; receive the QoE of an application included in the input information from an application function (AF) entity. In an implementation form of the first aspect, the network entity is or is included in a network data analytics function (NWDAF) entity, a PCF entity, or a SMF entity.

[0045] A third aspect of this disclosure provides a consumer network entity for applying a prioritization of packet types of packets of a QoS flow or of QoS flows of a service group, the consumer entity being configured to: receive output information of a network entity, wherein the output information comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow, or indicates one of the priority levels for each of two or more QoS flows of the service group; and apply, in the first future time period, the priority levels to the two or more packet types of the QoS flow, or to the two or more QoS flows of the service group.

[0046] The consumer network entity of the third aspect is able to apply the prioritization (respective priority levels) of the packet types of the QoS flow or QoS flows of the service group of QoS flows, which may have been determined considering dynamic network congestion levels reflecting network load and / or congestion. The applied prioritization may thus maximize the overall service capacity, and may retain sufficient user satisfaction under different network conditions. The consumer network entity of the third aspect may also apply updated priority levels. Notably, prioritizing packet types or QoS flows refers to applying the prioritizing strategy for these packet types or QoS flows, not necessarily to enforcing the prioritization. The enforcing may be done by another entity.

[0047] In an implementation form of the third aspect, the consumer network entity is further configured to provide a prioritization request to the network entity, the prioritization request indicating the QoS flow or the service group for which the prioritization is requested.

[0048] In an implementation form of the third aspect, the consumer network entity is further configured to: provide input information to the network entity, the input information comprising a number of priority levels and / or a network congestion level; and / or indicate the two or more packet types to the network entity.

[0049] A fourth aspect of this disclosure provides a method for determining a prioritization of packet types of packets of a QoS flow, the method comprising: receiving input information comprising a number of priority levels and / or a network congestion level; and providing, in a current time period, output information comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow.

[0050] The method of the fourth aspect may have implementation forms that correspond to the implementation forms of the network entity of the first aspect. Accordingly, the method of the fourth aspect may achieve all the advantages described above for the network entity of the first aspect, or any implementation form thereof.

[0051] A fifth aspect of this disclosure provides a method for determining a prioritization of QoS flows of a service group, the method comprising: receiving input information comprising a number of priority levels and / or a network congestion level; and providing, in a current time period, output information comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more QoS flows of the service group.

[0052] The method of the fifth aspect may have implementation forms that correspond to the implementation forms of the network entity of the second aspect. Accordingly, the method of the fifth aspect may achieve all the advantages described above for the network entity of the second aspect, or any implementation form thereof.

[0053] A sixth aspect of this disclosure provides a method for applying a prioritization of packet types of packets of a QoS flow, or of QoS flows of a service group, the method comprising: receiving output information of a network entity, wherein the output information comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow, or indicates one of the priority levels for each of two or more QoS flows of the service group; and applying, in the first future time period, the priority levels to the two or more packet types of the QoS flow, or to the two or more QoS flows of the service group.

[0054] The method of the sixth aspect may have implementation forms that correspond to the implementation forms of the consumer network entity of the third aspect. Accordingly, the method of the sixth aspect may achieve all the advantages described above for the consumer network entity of the third aspect, or any implementation form thereof.

[0055] A seventh aspect of this disclosure provides a computer program comprising instructions which, when the program is executed by a computer, cause the computer to perform the method according to the fourth aspect, fifth aspect, or sixth aspect of this disclosure.

[0056] An eighth aspect of this disclosure provides a non-transitory storage medium storing executable program code which, when executed by a processor, causes the method according to the third aspect, fourth aspect, or fifth aspect, or any of its implementation forms to be performed.

[0057] According to the above aspects and implementation forms, one solution provided by this disclosure is to leverage artificial intelligence (Al) and machine learning (ML) to prioritize the QoS fulfillment of certain packet types (e.g. MU types) or QoS flows in a service like an XR service in a mobile network. In particular, with the purpose to maximize the overall QoE under variable network load condition and also when the network is congested.

[0058] The present disclosure proposes the network entity of the first or second aspect (also referred to as QoS prioritizer in this disclosure), which may be Al and / or ML based, and is configured to derive the QoS prioritizing strategy (e.g. in QoS flow or packet (e.g., MU) granularity) in the mobile network.

[0059] The solutions of this disclosure include the definition of new network analytics (i.e., the prioritizing strategy, e.g., the first QoS priority setting) with the related input information and output information. The output information shows the priority level of each QoS flow or packet type, for example, in an XR service. For instance, in the form of a table, a list, or a formula at a certain time period. This information may be updated during the life time of a service. The input information may indicate a number of priority levels at different congestion level (e.g., via a table, a list, or a formula).

[0060] The disclosure also describes the AI / ML methods to derive the QoS prioritizing strategy. The related procedure for data collection, online / offline training, as well as the data inference in 3GPP system are presented. The procedure to consume the QoS prioritizing strategy in the 3GPP system is another part of the solution of this disclosure. It has to be noted that all devices, elements, units and means described in the present application could be implemented in the software or hardware elements or any kind of combination thereof. All steps which are performed by the various entities described in the present application as well as the functionalities described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. Even if, in the following description of specific embodiments, a specific functionality or step to be performed by external entities is not reflected in the description of a specific detailed element of that entity which performs that specific step or functionality, it should be clear for a skilled person that these methods and functionalities can be implemented in respective software or hardware elements, or any kind of combination thereof.

[0061] BRIEF DESCRIPTION OF DRAWINGS

[0062] The above described aspects and implementation forms will be explained in the following description of specific embodiments in relation to the enclosed drawings, in which

[0063] FIG. 1 shows an example of a dependent relationship in case of layered coding.

[0064] FIG. 2 shows an exemplary mapping of application layer packets to QoS flows in a mobile network.

[0065] FIG. 3 shows a network entity and a consumer network entity according to this disclosure.

[0066] FIG. 4 illustrates a solution of this disclosure, including the network entity (QoS prioritizer), the input information, and the output information consumed by the consumer network entity.

[0067] FIG. 5 shows an exemplary procedure for the QoS prioritizer of this disclosure, to collect the input information and to provide the output information, wherein the QoS prioritizer is implemented as a NWDAF entity. FIG. 6 shows exemplary AI / ML model training and inference methods that may be used by the QoS prioritizer of this disclosure.

[0068] FIG. 7 shows (a) an example of different sub-models of the QoS prioritizer of this disclosure, and (b) an example of a progressive priority level indication.

[0069] FIG. 8 shows (a) an exemplary operation in the QoS prioritizer of this disclosure, and (b) an exemplary procedure when the QoS prioritizer of this disclosure is implemented as an NWDAF entity.

[0070] FIG. 9 shows an example, in which the QoS prioritizer of this disclosure is used by a PCF entity (consumer network entity) for flow QoS parameter determination.

[0071] FIG. 10 shows an example, in which the QoS prioritizer of this disclosure is used by a SMF entity (consumer network entity) for flow QoS parameter determination.

[0072] FIG. 11 shows an exemplary deployment of the QoS prioritizer of this disclosure in a 3 GPP network architecture.

[0073] FIG. 12 shows (a) a QoS prioritizer of this disclosure implemented in a PCF entity, and (b) a QoS prioritizer of this disclosure implemented in an SMF entity.

[0074] FIG. 13 shows an exemplary mapping of an application layer priority to a network layer priority.

[0075] FIG. 14 shows a method for prioritizing packet types of packets of a QOS flow, according to this disclosure.

[0076] FIG. 15 shows a method for prioritizing QoS flows of a service group, according to this disclosure.

[0077] FIG. 16 shows a method for applying a prioritization of packet types of packets of a QoS flow, or of QoS flows of a service group, according to this disclosure. DETAILED DESCRIPTION OF EMBODIMENTS

[0078] FIG. 3 shows a network entity 300 according to this disclosure. The network entity 300 of FIG. 3 may correspond to the network entity of the first aspect or of the second aspect. The network entity 300 may be configured to determine a prioritization of packet types of packets 303 of a QoS flow 304, and / or may be configured to determine a prioritization of QoS flows 304 of a service group 305 of QoS flows 304. The network entity 300 may also be configured for both. The network entity 300 may be a network entity of a mobile network, for instance, a 5G or 6G mobile network. The network entity 300 may be or may be included in a NWDAF entity, or a PCF entity, or a SMF entity. The QoS flows 304 may be of an XR service. A packet 303 may be or correspond to a MU.

[0079] The network entity 300 is configured to receive input information 301, for instance, from one or more other entities of the mobile network. For instance, the network entity 300 may receive the input information 301 from an 0AM entity, a PCF entity, an SMF entity, and / or RAN entity. Different pieces of information composing the input information 301 may thereby be received from different entities. The input information 301 comprises a number of priority levels and / or comprises a network congestion level. The number of priority levels may, for example, be received from an 0AM entity. The network congestion level may, for example, be received from the 0AM entity, or from a RAN, or from an SMF entity. Further, the input information 301 may comprise a QoS fulfilment status for a current time period, wherein the QoS fulfilment status may be received from a PCF entity. The input information 301 may also comprise a current QoE of an application related to the QoS flow 304 or to the service group 305, respectively, wherein the QoE may be received from an AF entity.

[0080] The network entity 300 is further configured to provide, in the current time period, output information 302, for instance, to a consumer network entity 310. The output information 302 comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow 304, or indicates one of the priority levels for each of two or more QoS flows 304 of the service group 305. The first QoS priority setting is a QoS prioritization determined by the network entity 300. Accordingly, the network entity 300 is also referred to as a QoS prioritizer 300 in this disclosure. The consumer network entity 310 is configured to apply a prioritization of the packet types of the packets 303 of the QoS flow 304, or a prioritization of the QoS flows 304 of the service group 305. The consumer network entity 310 is adapted to receive the output information 302 of the network entity 300. Further, the consumer network entity 310 is configured to apply, in the first future time period, the priority levels of the first QoS priority setting to the two or more packet types of the QoS flow 304, or to the two or more QoS flows 304 of the service group 305, as illustrated in FIG. 3. The consumer network entity 310 may apply the priority levels by outputting a respective instruction or message 311, for instance, to be received and used by another entity for enforcing the priority levels.

[0081] The network entity 300 and / or the consumer network entity 310 may each comprise a processor or processing circuitry (not shown) configured to perform, conduct or initiate the various operations of the network entity 300 and / or the consumer network entity 310 described herein. The processing circuitry may comprise hardware and / or the processing circuitry may be controlled by software. The hardware may comprise analog circuitry or digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as applicationspecific integrated circuits (ASICs), field-programmable arrays (FPGAs), digital signal processors (DSPs), or multi-purpose processors. The network entity 300 and / or the consumer network entity 310 may further each comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under control of the software. For instance, the memory circuitry may comprise a non-transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the network entity 300 and / or the consumer network entity 310 to be performed. In one embodiment, the processing circuitry comprises one or more processors and a non-transitory memory connected to the one or more processors. The non-transitory memory may carry executable program code which, when executed by the one or more processors, causes the network entity 300 and / or the consumer network entity 310 to perform, conduct or initiate the operations or methods described herein.

[0082] FIG. 4 illustrates a QoS prioritizer 300 of this disclosure, which builds on the network entity 300 shown in FIG. 3, and illustrates examples of pieces of information composing the input information 301, and the output information 302 consumed by the consumer network entity 310 of this disclosure. In case of prioritizing QoS flows 304 of a service group 304, for example, QoS flows 304 of an XRM service, the input information 301 and the output information 302 of the QoS prioritizer 300 may be as follows.

[0083] The input information 301 may comprise a QoE 402 of the application, flow IDs or a service group ID, QoS monitoring results (e.g., QoS fulfillment status 403 like QoS fulfilled / not fulfilled, PER, PDB fulfilled / not fulfilled, PSDB, PSER fulfilled / not fulfilled), and current priority settings of each QoS flow 304 in the service group 305, and optionally an application type. The input information 301 further comprise the number of priority levels 401 (for example, priority levels 401 per network load, as illustrated), and may comprise an indication of a network load 404. The various pieces of information 401, 402, 403, and 404 composing the input information 301 may be received by the QoS prioritizer 300 from different entities, like from AF entity, PCF entity, 0AM entity, RAN entity, and / or SMF entity, for instance, as illustrated in FIG. 4.

[0084] The output information 303 comprises the first QoS priority setting, for example, a priority level setting per QoS flow 304 in the service group 305 and the related time period. The output information 302 may comprise priority level setting(s) per QoS parameter per QoS flow 304 in the service group 305 and the related time period.

[0085] In this case, the service group ID indicates a group of QoS flows 304 belonging to the same XRM service. The service group ID is also named flow group ID in this disclosure. The related time period refers to the time period when the output priority setting(s) apply.

[0086] In case of prioritizing the packet types of the packets 303 (e.g., MUs) of the QoS flow, for example, a QoS flow 304 of an XRM service, the input information 301 and output information 302 of the QoS prioritizer 300 may be as follows.

[0087] The input information 301 may comprise a QoE 402 of the application, packet types, MU type(s), service group ID, flow QoS monitoring results (e.g., QoS fulfillment status 403 like QoS fulfilled / not fulfilled, PER, PDB fulfilled / not fulfilled, PSDB, PSER fulfilled / not fulfilled) and current priority settings of each packet type in the QoS flow 304, and optionally an application type. The input information 301 further comprise the number of priority levels 401 (for example, a priority level 401 per network load as illustrated), and may comprise an indication of a network load 404. The various pieces of information 401, 402, 403, and 404 composing the input information 301 may be received from different entities, like AF entity, PCF entity, 0AM entity, RAN entity, and / or SMF entity, for instance, as illustrated in FIG. 4.

[0088] The output information 303 comprises the first QoS setting, for example, a priority level setting per packet type of the packets 303 in the QoS flow 304 and the related time period. The output information 302 may comprise the priority level setting(s) per MU type and / or per MU position (e.g., first P-frame after an I-frame) and the related time period.

[0089] As specifically illustrated in FIG. 4, in each of the two above-described cases, the input information 301 may comprise a respective number of priority levels 401 for each of two or more network congestion levels 404. Further, the input information 301 may comprise the current QoE 402 of the application related to the QoS flow 304 or to the service group 305. The QoS prioritizer 300 may be configured to obtain a relationship between the first QoS priority setting for the first future time period and the QoE 402 during the first future time period, and to determine and provide the output information 302 for a second future time period (later than the first future time period) based on the relationship and based on an award to increase the QoE 402.

[0090] The QoS prioritizer 300 may also be configured to determine the first QoS priority setting according to the network congestion level 404 and / or according to the determined QoS fulfilment status 403, in case these pieces of information are included in the input information

[0091] 301. As illustrated in FIG. 4, the QoS prioritizer 300 may receive these pieces of information 403 and 404 for the current time period by indication from another network entity, like PCF entity and RAN or SMF or 0AM entity. However, the network entity 300 could also be configured to determine the network congestion level 404 and / or the QoS fulfilment status 403 for the first future time period based on the network congestion level 404 and / or QoS fulfilment status 403 for the current time period.

[0092] FIG. 5 shows an exemplary procedure for the QoS prioritizer 300 of this disclosure, wherein the procedure is for collecting the input information 301 and providing the output information

[0093] 302, and wherein the QoS prioritizer 300 is implemented as an NWDAF entity. The procedure is as follows. 1. A consumer network entity 310 of the QoS prioritizing strategy (e.g., an NF like a PCF or SMF entity) identifies an XRM service using flow group ID or flow ID(s), and subscribes to the prioritizing strategy analytics service in the NWDAF implementing the QoS prioritizer 300. The consumer network entity 310 may optionally indicate to use progressive prioritizing.

[0094] 2. The QoS prioritizer 300 in the NWDAF collects the QoS fulfillment status 403 and priority settings of the related QoS flows 304, packet types (e.g., MU types), UPF load / flow or session congestion level, etc. from the network functions. The priority settings may be included as QoS parameters in the related QoS policy of a flow or MU. The QoS prioritizer 300 in the NWDAF collects the QoE information 402 from the application via the AF. The QoS prioritizer 300 in the NWDAF obtains QoS priority levels 401 per network load condition / congestion level from the 0AM entity or other NFs.

[0095] 3. The QoS prioritizer 300 in the NWDAF derives the priority level for a QoS flow 304 or a packet type at a certain time period. This step may include multiple phases as detailed further below.

[0096] 4. The QoS prioritizer 300 in the NWDAF notifies the prioritizing strategy (first QoS priority setting) to the consumer NF 310 per time period or when the prioritizing strategy changes.

[0097] FIG. 6 shows an example of AI / ML model training and inference methods, which may be used by the QoS prioritizer 300 of this disclosure.

[0098] The QoS prioritizer 300 may utilize AI / ML technology to derive the QoS prioritizing strategy. In an example implementation, the ML model uses a NN, e.g., combined with a reinforcement learning approach. The NN may be used to learn the relationship between the dynamic QoS priority setting and the QoE 402. The reinforcement learning using QoE feedback as a reward may constantly improve the ML model of the QoS prioritizer 300.

[0099] FIG. 6 specifically shows how the QoS prioritizer 300 can use the reinforcement learning. State information collected from the environment may include the QoS policy, the QoS fulfillment status 403, or the like, and the reward is the QoE 402. The action is the prioritizing strategy (i.e., the derived QoS priority levels 401 of a QoS flow 304 or packet type in the network). The operation of QoS prioritizer 300 may comprise an offline ML model training phase, an online ML model training phase, and an inference phase (to derive the analytics with the input information 301).

[0100] During the offline training phase, as illustrated in FIG. 7(a) and as described above, a ML model

[0101] 701 may be trained for the QoS prioritizer 300 to use. Alternatively, instead of a single model 701, a set of one or more sub-models 702 may be trained for the QoS prioritizer 300 to use. Each sub-model 702 may represent a differently trained version of the model 701, that is, each sub-model 702 may result from different training (e.g., with different training data and / or Al configuration) of the model- 701. Each sub-model 702 may also be regarded as an updated model 701, e.g. updated by the QoS prioritizer if the offline training is performed locally. The different sub-models 702 may differ from each other with respect to the number of priority levels 401 they consider and are trained for. While the simplest sub-model 702 may consider only two priority levels 401 (one higher, one lower), more complex sub-models 702 can retrieve a number of priority levels that is arbitrarily high. The set of trained sub-models 702 can only be deployed, if all are capable of achieving a sufficient level of accuracy.

[0102] The QoS prioritizer 300 may initially (i.e., as an initial mode) use the model 701 or one or more of the sub-models 702, which may have respectively been trained offline for similar application, until a sufficient level of accuracy can be achieved. The initial model(s) 701, 702 can then be further trained (online) using the collected data for the similar XRM service(s). The model(s) 701, 702 might be constantly improved using more data and the QoE feedback from similar XRM services during the online training phase. Notably, the sub-models 702 may be trained (offline or online) using a different number of priority level settings (e.g., 2 levels of priority, n levels of priority).

[0103] The QoS prioritizer 300 may also iteratively use different sub-models 702. For instance, the QoS prioritizer 300 can use a first sub-model 702 as initial model in a first iteration, may optionally train the first sub-model 702 using online data, and may then use the first sub-model

[0104] 702 for inference. In a second iteration, the QoS prioritizer 300 may use a second sub-model 702 as an initial model, may (optionally) train the second sub-model 702 with online data, and may then use the second sub-model 702 for inference. During the offline training (for building the initial model(s) 701, 702), the model(s) 701, 702 may learn all the key relationships between the input features and the model’s output. Hence, the offline training phase may be beneficial for building the model(s) 701, 702, while the online training phase can take optional inputs for further improvements of the model(s) 701, 702 (i.e., refining already known relationships / correlations learned during offline training).

[0105] In the inference phase, the QoS prioritizer 300 may use the collected data from the current XRM service (e.g., MU types of the current XRM service, congestion level of the service flow of the current XRM service, network condition, application ID, etc.) for analytics generation. During the inference phase, the QoE may not considered as the input. Here the MU types can be different type of frames / slice (I, P), Instantaneous Decoder Refresh (IDR)), in / out FOV, video layers (e.g., base layer, enhancement layer), etc.

[0106] During the inference phase, QoE feedback can be provided, allowing to further improve the inference accuracy. For improvement, there are two options: 1) use another offline trained model 701, or 702) improve the current model 701, 702. In both cases, the QoE feedback may determine how effective the new / updated model is. In both cases, this may be done as follows.

[0107] Available is a set of sub-models 702 (either the model 701 and its potentially improved version(s) (through online training) or the set of sub-models 702 trained during the offline training phase). The sub-models 702 can differ in terms of the number of priority levels they consider. The ultimate goal is to select the most appropriate sub-model 702. For instance, the sub-model 702 to select may be the one providing the best QoE improvement, as compared to a baseline (no intelligent QoS prioritization strategy) or the previous sub-model 702 (which was subject of improvement). The QoE improvement which a certain sub-model 702 achieves may be assessed using the QoE feedback. Therefore, QoS flows 304 which have similar network and application characteristics (e.g., the same network congestion and using the application) may be treated according to the output given by the different sub-models 702 under investigation. As the QoS flows 304, to which the different sub-models 702 are applied, have the same characteristics, the QoE feedback allows to determine the most effective sub-model 702. The most effective sub-model 702 is the one to be deployed and potentially further improved during the online training phase. The online training phase is optional and only applied if the respective additional input information 301 (e.g. the QoE 402) is provided. While with online training, allows the model 701 (besides improving its output) to adapt to dynamics in the XR application (e.g. increased number of video quality levels over time).

[0108] Since the ML model 701 may not be very accurate in the beginning of the life time of a QoS flow 304 or service group 305 of QoS flows 304, the QoS prioritizer 300 may be configured with a progressive priority level indication 703, as shown in FIG. 7(b). For instance, this may indicate to use less number of QoS priority levels in the beginning of the life time of a QoS flow 304 or service group 305 of QoS flows 304, and to increase the QoS priority levels gradually (more accurate) in a later phase. Here the additional input of the QoS prioritizer 300 may be an indication for progressive priority level. Again here the two options mentioned above apply, i.e., (1) use another offline trained sub-model 702 with increased number of QoS priority levels, or (2) improve the current ML model 701 with a different configuration of increased number of QoS priority levels.

[0109] The detailed operation of the QoS prioritizer 300 as indicted above then includes also the offline ML model training phase and online ML model training phase and inference phase. In all the phases, the input information 301 and output information 302 may be the same as defined previously, for instance, with respect to FIG. 4, except for the QoE 402 which may not be used in the inference phase. The input information 301 may be collected from the NFs using the procedure as shown in FIG. 5 and the output information 302 may be provided to the consumer network entity 302 using the procedure shown in FIG. 5.

[0110] An exemplary related operation is illustrated as a flow chart in FIG. 8(a). And an exemplary procedure when the QoS prioritizer 300 is implemented as a NWDAF is illustrated as message sequence chart in FIG. 8(b). The Al configuration refers to the configuration used for the ML model training such as the Al algorithm and the additional configurations such as the constraints (e.g., number of priority levels). The online training may occur in parallel to the analytics inference (e.g., improve the ML model 701 constantly during the analytics generation). In case progressive priority indication 703 is on, step 3 and step 4 will be performed recursively. The QoS prioritizer 300 may increase the number of priority levels after certain accuracy is reached in each iteration of the training and inference until the number of priority levels in the Al configuration. Alternatively, the QoS prioritizer 300 may use an offline trained sub-model 702 with increased number of priority levels directly in different iteration of the inference instead of performing online training with different configuration.

[0111] FIG. 9 and FIG. 10 illustrate exemplary use cases of the QoS priority setting (in the output information 302 of the QoS prioritizer 300).

[0112] As shown in FIG. 9, the PCF entity as the consumer network entity 310 may subscribe to the priority strategy for QoS flows 304 in a XRM service provided by the QoS prioritizer 300. The obtained priority level(s) can be used to set the priority / ARP level of the correspondent QoS flow 304. The flow priority or ARP level may be provided as part of the QoS parameter with the traffic description of that flow (i.e., 5 tuples) to the SMF entity using policy association procedure. The SMF entity determines the QoS profile for that QoS flow 304 accordingly and configures the UPF entity using the N4 procedure for the related QoS treatment. The SMF entity provides the QoS profile (including the flow priority / ARP) together with the traffic description to the RAN entity using the PDU session modification procedure. The RAN entity may be able to perform related QoS treatment based on the flow priority / ARP. The related treatment includes prioritized flow scheduling, prioritized flow resource preemption.

[0113] As shown in FIG. 10, the SMF entity as the consumer network entity 310 may subscribe to the priority strategy for different packet types (importance) in XRM service, as provided by the QoS prioritizer 300. The obtained priority level can be used to set the packet detection and marking rules in the UPF entity. The SMF entity provides the packet detection rules and packet marking rules to the first UPF entity (where the service traffic enters the 3 GPP network) as configuration information using the N4 procedure. The UPF entity performs packet detection (e.g., based on detected packet type / importance information in the packet header) and marks the priority level in the network layer of the packet (e.g., GTP header). Then packets with marked GTP header are transported to RAN entity in the user plane. The RAN entity is able to read the GTP header to understand the priority level of, for instance, the MU / PDU set. Then the RAN entity may be able to perform related QoS treatment based on the priority level(s). The related treatment includes prioritized packet scheduling, prioritized packet dropping, etc.

[0114] FIG. 11 shows an exemplary system architecture or scenario, to which the solutions of the present disclosure are applicable. FIG. 11 specifically shows an example of the QoS prioritizer 300 deployed in a 3GPP network as an NWDAF entity. Based on the analytics subscription from a consumer network entity 310 (e.g., SMF entity), the QoS prioritizer 300 collects the input information 301 from the related network functions, runs an AI / ML algorithm based on the collected information for the ML model training / retraining and analytics inference. The result analytics is provided to the consumer network entity 310 as the QoS prioritizing strategy (e.g., mapping of the packet type(s) to priority level(s)). The consumer network entity 310 uses the obtained priority level(s) for the QoS control (e.g., uses in the related PCC rule or UPF configurations). The priority level used in the related PCC rule is distributed in the 3GPP system using the PCC rule association and distribution procedure. Then SMF / RAN entity is able to perform related prioritized QoS treatment accordingly. The priority levels used in the packet detection and marking rule is distributed in the 3GPP system using the N4 configuration procedure and UP transport. Then the UPF / RAN entity is able to perform the related prioritized QoS treatment accordingly.

[0115] The present disclosure relates to the NWDAF, PCF, SMF, for instance, of a 5G or 6G mobile network, wherein the QoS prioritizer 300 may be implemented as NWDAF, or inside PCF, or inside SMF. The input information 301 and the output information 302 are reflected in related messages and procedures as shown in FIG. 12 and FIG. 13. The UPF entity may also be enhanced to be able to perform packet marking based on the prioritizing strategy provided by the QoS prioritizer 300 (e.g., a mapping table instead of read and copy directly to the GTP header) from the SMF entity. Possible implementation forms of the present disclosure can be computer programs in hardware / software implementing the proposed AI / ML processing logic in the NWDAF, PCF or SMF.

[0116] FIG. 12 shows the QoS prioritizer 300 exemplarily implemented in an NF. The QoS prioritizer 300 can also be implemented as part of an existing network function, e.g. (1) as a part of a PCF entity to prioritize the QoS flows 304 in an XRM service (FIG. 12(a)), and (2) as a part of the SMF entity to prioritize MUs in an XRM service (FIG. 12(b)). Then, the PCF / SMF entity instead of the NWDAF entity may subscribe to the related AF / NF events (to collect the input information 301) in this case.

[0117] Similar data may be collected compared to the case where the QoS prioritizer 300 is implemented as NWDAF, while maybe from a different network function or even locally available in the network function (e.g., the QoS priority settings). The derived prioritization strategy may be embedded directly in the related control message (e.g., PCC rule in case of the QoS prioritizer 300 being implemented in the PCF entity, N4 configuration in case of the QoS prioritizer 300 implemented in the SMF entity) instead of an analytics result. The PCC rule procedure and N4 configuration procedure may apply here for the distribution of the derived QoS prioritize strategy. The number of priority levels per network load can also be provided by the NWDAF entity based on some analytics instead of directly from the 0AM entity. When the QoS prioritizer 300 is implemented in a NF, the QoS prioritizer 300 may also get the flow ID(s) from the AF / PCF entity, since that is not in the subscription information of the analytics ID from the consumer NF any more.

[0118] FIG. 13 shows an exemplary mapping of an application layer priority to a network priority. The QoS prioritizer 300 may get application layer importance information per QoS flow and / or packet type directly from the application. Such information can be partial or complete, e.g., only the most important flow in a flow group, or only the importance of the I-frame.

[0119] Then the QoS prioritizer 300 can be used to derive the missing priority information, or squeeze more priority levels into the desired number of priority levels under a certain network condition (as shown in FIG. 13).

[0120] FIG. 14 shows a method 1400 for determining a prioritization of packet types of packets 303 of a QoS flow 304. The method 1400 may be performed by a network entity 300 of the first aspect. The method 1400 comprises a step 1401 of receiving input information 301 comprising a number of priority levels and / or a network congestion level. Further, the method 1400 comprises a step 1402 of providing, in a current time period, output information 302 comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow 304.

[0121] FIG. 15 shows a method 1500 for determining a prioritization of QoS flows 304 of a service group 305. The method 1500 may be performed by a network entity 300 of the second aspect. The method 1500 comprises a step 1501 of receiving input information 301 comprising a number of priority levels and / or a network congestion level. Further, the method 1500 comprises a step 1502 of providing, in a current time period, output information 302 comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more QoS flows 304 of the service group 305.

[0122] FIG. 16 shows a method 1600 for applying a prioritization of packet types of packets of a QoS flow 304 or of QoS flows 304 of a service group 305. The method 1600 may be performed by a consumer network entity 310. The method 1600 comprises a step 1601 of receiving output information 302 of a network entity 300, wherein the output information 302 comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels for each of two or more packet types of the QoS flow 304, or indicates one of the priority levels for each of two or more QoS flows 304 of the service group 305. The method 1600 further comprises a step 1602 of applying, in the first future time period, the priority levels to the two or more packet types of the QoS flow 304, or to the two or more QoS flows 304 of the service group 305.

[0123] Some advantages achieved by the solution of the present disclosure are as follows. The Al based QoS prioritizing based on the QoE from the application allows omitting that the application provides the (complete) priority information to the network, resulting in less complexity and better network performance. In particular, this leads to less complexity for the application to provide such information, and avoids the need for additional network processing to obtain such information (e.g., packet inspection). It also makes the solution applicable for encrypted service traffic.

[0124] The Al based QoS prioritizing using as constraint the number of priority levels based on network load condition enables controlling the QoS strategy jointly, to consider the prioritizing overhead in the network under different network loads. This results in higher flexibility, and better resource efficiency. For example, this enables to balance the trade off by controlling the number of priority levels in the network. Further, a stable number of priority levels may be maintained in the network despite the dynamicity of priority levels, leading to better resource efficiency. For instance, a GOP structure (number of priority levels) may change depending on the scene, while very dynamic priority implementation in the life time of a service flow is inefficient in the network. The AI / ML model training methods with offline trained initial ML model and progress inference with increased priority level allow for fast determination of the prioritize strategy and ensure smooth system operation even if the AI / M1 model is not sufficiently trained yet. The method to use the prioritize strategy (by QoS prioritizer 300 / analytics consumer) in the 5G+ mobile network enables using the method in mobile networks with compatible 3GPP procedures.

[0125] The present disclosure has been described in conjunction with various embodiments as examples as well as implementations. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed matter, from the studies of the drawings, this disclosure and the independent claims. In the claims as well as in the description the word “comprising” does not exclude other elements or steps and the indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation.

Claims

CLAIMS1. A network entity (300) for determining a prioritization of packet types of packets (303) of a quality of service, QoS, flow (304), the network entity (300) being configured to: receive input information (301) comprising a number of priority levels (401) and / or a network congestion level (404); and provide, in a current time period, output information (302) comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more packet types of the QoS flow (304).

2. A network entity (300) for determining a prioritization of QoS flows (304) of a service group (305), the network entity (300) being configured to: receive input information (301) comprising a number of priority levels (401) and / or a network congestion level (404); and provide, in a current time period, output information (302) comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more QoS flows (304) of the service group (305).

3. The network entity (300) according to claim 1 or 2, further configured to determine the first QoS priority setting based on the received number of priority levels (401) and / or the received network congestion level (404).

4. The network entity (300) according to claim 3, wherein: the input information (301) comprises a respective number of priority levels (401) for each of two or more network congestion levels (404).

5. The network entity (300) according to claim 4, further configured to: determine a network congestion level (404) and / or a QoS fulfilment status (403) for the first future time period; and determine the first QoS priority setting according to the determined network congestion level (404) and / or the determined QoS fulfilment status (403).

6. The network entity (300) according to claim 5, configured to: receive a network congestion level (404) and / or a QoS fulfilment status (403) for the current time period by indication from another network entity; and determine the network congestion level (404) and / or the QoS fulfilment status (403) for the first future time period based on the network congestion level (404) and / or QoS fulfilment status (403) for the current time period.

7. The network entity (300) according to one of the claims 1 to 6, wherein the input information (301) further comprises at least one of an application ID; a flow ID or flow description; a current quality of experience, QoE, (402) of an application related to the QoS flow (304) or to the service group (305); a second QoS priority setting for the current time period, wherein the second QoS priority setting indicates one of the priority levels (401) for each of the two or more packet types of the QoS flow (304) or for each of the two or more QoS flows (304) of the service group (305); an indication of one or more packet types of the QoS flow (304): a service group ID of the service group (305); a QoS fulfilment status (403) for the current time period, wherein the QoS fulfilment status (403) is related to one or a combination of QoS parameters of a QoS flow (304) selected from: a packet error rate, a packet delay budget, a protocol data unit set error rate, and / or a protocol data unit set delay budget, of the packets (303) of the QoS flow (304) or of the QoS flows (304) of the service group (305).

8. The network entity (300) according to one of the claims 1 to 7, further configured to receive application layer priority information indicating the two or more packet types.

9. The network entity (300) according to one of the claims 1 to 8, configured to repeatedly update and provide the output information (302) every time period or every time the output information (302) changes.

10. The network entity (300) according to one of the claims 1 to 9, configured to determine the output information (302) using a trainable model (701, 702).

11. The network entity (300) according to claim 10, configured to obtain the trainable model (701, 702) from another network entity and / or train the trainable model (701, 702) based on the input information (301).

12. The network entity (300) according to one of the claims 1 to 11, configured to: obtain a relationship between the first QoS priority setting for the first future time period and a QoE (402) during the first future time period of an application related to the QoS flow (304) or to the service group (305); and determine and provide the output information (302) for a second future time period based on the relationship and based on an award to increase the QoE (402).

13. The network entity (300) according to one of the claims 10 to 12, wherein the trainable model (701, 702) comprises a plurality of sub-models (702), wherein each sub-model (702) is configured for a different number of priority levels (401); and the network entity (300) is configured to select one of the sub-models (702) based on the number of priority levels (401) included in the input information (301), and to use the selected sub-model (702) for determining the output information (302).

14. The network entity (300) according to claim 13, further configured to select the one of the sub-models (702) based on the relationship and based on the award to increase the QoE (402).

15. The network entity (300) according to claim 13 or 14, wherein: the input information (301) further comprises a progressive priority level indication (703), and according to the progressive priority level indication (703), the network entity (300) is configured to update at least one of the sub-models (702) by stepwise increasing the number of priority levels (401) related to the sub-model (702); or the network entity (300) is configured to select the one of the sub-models (702) based further on an increased number of priority levels (401).

16. The network entity (300) according to one of the claims 1 to 15, configured to provide the output information (302) to a consumer network entity (310), for example, a policy control function, PCF, entity or a session management function, SMF, entity.

17. The network entity (300) according to one of the claims 3 to 16, configured to: receive the number of priority levels (401) or a respective number of priority levels (401) for each of two or more network congestion levels (404) included in the input information (301) from an 0AM entity; and / or receive at least one of the second QoS priority setting, the QoS fulfilment status (403), and the network congestion level (404) included in the input information (301) from a PCF entity, or a SMF entity, or a radio access network, RAN, entity; receive the QoE (402) of an application included in the input information (301) from an application function, AF, entity.

18. The network entity (300) according to one of the claims 1 to 17, wherein the network entity (300) is or is included in a network data analytics function, NWDAF, entity, a PCF entity , or a SMF entity.

19. A consumer network entity (310) for applying a prioritization of packet types of packets (303) of a quality of service, QoS, flow (304) or of QoS flows (304) of a service group (305), the consumer entity (310) being configured to: receive output information (302) of a network entity (300), wherein the output information (302) comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more packet types of the QoS flow (304), or indicates one of the priority levels (401) for each of two or more QoS flows (304) of the service group (305); and apply (311), in the first future time period, the priority levels (401) to the two or more packet types of the QoS flow (304), or to the two or more QoS flows (304) of the service group (305).

20. The consumer network entity (310) according to claim 19, further configured to: provide a prioritization request to the network entity (300), the prioritization request indicating the QoS flow (304) or the service group (305) for which the prioritization is requested.

21. The consumer network entity (310) according to claim 19 or 20, further configured to:provide input information (301) to the network entity (300), the input information (301) comprising a number of priority levels (401) and / or a network congestion level (404); and / or indicate the two or more packet types to the network entity (300).

22. A method (1400) for determining a prioritization of packet types of packets (303) of a quality of service, QoS, flow (304), the method (1400) comprising: receiving (1401) input information (301) comprising a number of priority levels (401) and / or a network congestion level (404); and providing (1402), in a current time period, output information (302) comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more packet types of the QoS flow (304).

23. A method (1500) for determining a prioritization of QoS flows (304) of a service group (305), the method (1500) comprising: receiving (1501) input information (301) comprising a number of priority levels (401) and / or a network congestion level (404); and providing (1502), in a current time period, output information (302) comprising a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more QoS flows (304) of the service group (305).

24. A method (1600) for applying a prioritization of packet types of packets (303) of a quality of service, QoS, flow (304) or of QoS flows (304) of a service group (305), the method (1600) comprising: receiving (1601) output information (302) of a network entity (300), wherein the output information (302) comprises a first QoS priority setting for a first future time period, wherein the first QoS priority setting indicates one of the priority levels (401) for each of two or more packet types of the QoS flow (305), or indicates one of the priority levels (401) for each of two or more QoS flows (304) of the service group (305); and applying (1602), in the first future time period, the priority levels (401) to the two or more packet types of the QoS flow (304), or to the two or more QoS flows (304) of the service group (305).

25. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to perform the method (1400, 1500, 1600) according to one of the claims 22 to 24.

Images