Techniques for selecting a network function entity based on resource and time information
Patent Information
- Application Number
- PCT/EP2025/054747
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-02-21
- Publication Date
- 2026-08-27
Smart Images
Figure EP2025054747_27082026_PF_FP_ABST
Abstract
Description
[0001] TECHNIQUES FOR SELECTING A NETWORK FUNCTION ENTITY BASED ON RESOURCE AND TIME INFORMATION
[0002] TECHNICAL FIELD
[0003] The disclosure relates to the field of radio access networks, in particular next generation radio access networks, particularly for 5G systems. The disclosure relates to a network function selection entity and techniques for selecting a network function entity based on resource and time information, and corresponding network function entities. In particular, the disclosure relates to optimal NF selection based on network constraints parameters.
[0004] BACKGROUND
[0005] Network function selection in accordance with 3GPP specifications is pivotal for the effective operation of modern mobile networks. By ensuring that the right network functions are selected to handle user requests based on a variety of factors and parameters (such as network resource parameters, e.g. computation time, energy consumption, etc..), mobile network operators can enhance service quality, improve resource efficiency, and deliver a better overall experience to users. As the industry continues to evolve, ongoing updates and refinements to these selection processes will be essential in adapting to new technological advancements and user demands.
[0006] SUMMARY
[0007] This disclosure provides a new robust and efficient concept of Network Function selection for the 5G network taken into account the scalability, real time application, compatibility and resource overhead.
[0008] The foregoing and other objects are achieved by the features of the independent claims. Further implementation forms are apparent from the dependent claims, the description and the figures.
[0009] NF selection in 5G networks is crucial for optimizing the use of network resources and ensuring efficient performance. The disclosure presents different embodiments to consider regarding NF selection while taking network resources into account: 1 )-Load Balancing: Distributing the load evenly across NFs helps prevent bottlenecks and makes better use of available resources. Dynamically adjust NF selection based on real-time traffic conditions and resource utilization. 2)-Scalability: The ability to scale NFs up or down based on demand is critical for effective resource utilization. Select NFs that can easily be scaled horizontally or vertically without significant overhead. 3)- Quality of Service (QoS): Ensure that selected NFs can meet the QoS requirements of different service types, taking into account the available resources. Monitor and adjust NF selection based on QoS metrics. 4)-Energy Efficiency: Optimize NF selection to reduce energy consumption in the network, especially as demand fluctuates. Use energy-aware algorithms to inform NF placement and resource allocation decisions.
[0010] The solution presented in this disclosure provides an effective NF selection in 5G networks which involves a comprehensive approach that balances performance, resource constraints, and user demands to optimize the network's overall functioning.
[0011] For task execution, embodiments of the disclosure take into account two main metrics of 5G Advanced network, which are required Resource (Rf) and required Time (Tf). Those parameters are computed locally by NF.These two metrics, required Resource (Rf) and required Time (Tf) may be determined, for example, according to the following:
[0012] Ri(normalized) is the required resource (such as required energy consumption, computation and communication overheads) that network function needs to execute a specific service, e.g., training, event detection and classification process, NF selection, etc.
[0013] Tfis the required time (normalized) of network function on executing the request task. Tfmay be for example close to one when at iteration t, NF spends a high task execution time as compared to the previous iterations t-K(K is the total number of iterations). Tfmay be for example close to zero when at iteration t, NF spends a very low task execution time as compared to the previous iterations t-K.
[0014] Note that the above definitions are only one exemplary way to express Riand Ti, other definitions may be applied as well.
[0015] Any service / service operation depends on the required Resource (Ri) and required time (Ti) that NF needs to execute it.
[0016] Other parameters related to priority / weight factors for load balancing, service chaining, flexibility and adaptability and quality of experience may also be considered.
[0017] In order to describe the disclosure in detail, the following terms and notations will be used.
[0018] Federated Learning FL
[0019] Vertical Federated Learning VFL
[0020] Network Data Analytics Function NWDAF
[0021] Application function AF
[0022] Network Function NF
[0023] Network repository function NRF
[0024] NF service consumer NFc
[0025] NF service producer NFp
[0026] Network exposure function NEF
[0027] Public land mobile network PLMN
[0028] Model training logical function MTLF
[0029] Policy and Charging Control PCC
[0030] Policy Control Function PCF
[0031] Service Communication Proxy SCP
[0032] In this disclosure, network function selection is described. The Network Function (NF) selection is a critical aspect of the 3GPP specifications, particularly for 5G systems. The 3GPP framework defines a set of procedures and protocols to ensure efficient and reliable communication between various NF deployed in the 5G network. The NF selection process involves several steps: 1)-NF Registration: Network functions register their profiles with the Network Repository Function (NRF). This includes information about the NF's capabilities and services. 2)-NF Discovery: When a NF needs to interact with another NF, it sends a discovery request to the NRF. The NRF then provides information about available NF instances that match the requested criteria. 3)-NF Selection: The requesting NF selects an appropriate NF instance based on the information provided by the NRF. This selection can be influenced by factors such as load balancing, proximity, and specific service requirements. 4)-NF Communication: Once the NF instance is selected, communication is established using the defined service-based interfaces. In practice, NF selection ensures that network functions can efficiently discover andcommunicate with each other, enabling seamless and reliable service delivery to end-users. This is particularly important in scenarios where network resources need to be dynamically allocated and managed to meet varying demands. By adhering to the 3GPP specifications, network operators can implement robust and scalable NF selection mechanisms, contributing to the overall performance and reliability of the 5G network.
[0033] According to a first aspect, the disclosure relates to a network function selection entity for selecting a network function entity of a plurality of network function entities, the network function selection entity comprising: a receiver configured to collect information about a plurality of network function entities which are capable of providing a service in a communication network, the service being identified by a service identifier; wherein the information comprises information about a resource required by a respective network function entity of the plurality of network function entities to execute the service; and information about a time required by the network function entity to execute the service; and a processor configured to select at least one network function entity of the plurality of network function entities based on the collected information about resources and times of the network function entities which resources and times are matching a threshold criterion.
[0034] Such network function selection entity provides a new robust and efficient mechanism for Network Function selection in the 5G network. The network function selection entity enables scalability, real time application, compatibility and resource overhead.
[0035] The resource / time information can be received from the network function (NF) entity but also be retrieved using an OAM process, or any other mechanism. For instance, the NF selection entity can receive the information from another NF, the NWDAF (see Figure 2). The goal is that this information is made available to the selecting NF, i.e., the network function selection entity.
[0036] In an exemplary implementation of the network function selection entity, the threshold criterion is based on at least one of the following: load balancing for an evenly load distribution between the network function entities; scalability to scale the network function entity up and down without significant overhead; quality of Service, QoS, to ensure that the selected network function entity can meet QoS requirements of different service types; energy efficiency to reduce energy consumption in the communication network.
[0037] The network function selection entity thus provides a flexible load balancing scalability, QoS and reduced energy consumption.
[0038] In an exemplary implementation of the network function selection entity, the information about a resource comprises: a required energy consumption for executing the service; a required computation complexity for executing the service; a required communication overhead for executing the service.
[0039] The network function selection entity can thus flexibly select an optimal network function entity that meets the above requirements.
[0040] In an exemplary implementation of the network function selection entity, the processor is configured to select at least one network function entity which resource required to execute a service is below a resource threshold and / or which time required to execute a service is below a time threshold.
[0041] The idea here is to select a most efficient function, i.e. low resource consumption and / or low response / processing time.In an exemplary implementation of the network function selection entity, the resource threshold is based on at least one of the following: compute resources availability on one or more edge servers or in a cloud infrastructure; network bandwidth for the service; storage capacity; load balancing; CPU usage; memory usage; disk usage; data volume; and the time threshold is based on at least one of the following: latency requirements; response time of the network function entity; processing time of the network function entity; orchestration timing for performing load balancing or scaling actions.
[0042] These different parameters can be exploited for selecting the optimal network function entity.
[0043] In an exemplary implementation of the network function selection entity as shown in Figure 2, for example, the network function selection entity is a network function service consumer, NFc, entity of a 5G network which is configured to receive the information about the plurality of network function entities from a Network Data Analytics Function, NWDAF, entity of the 5G network.
[0044] The network function selection is most suitable for the 5G Advanced context, as it considers both 5G Advanced metrics, R and T during the task execution process. The operator is enabled to implement different network function trade-offs, depending on how critical the 5G Advanced task is and depending on the T and R requirements associated with this task.
[0045] In an exemplary implementation of the network function selection entity, the NFc entity is configured to request the information about the plurality of network function entities from the NWDAF entity.
[0046] The NFc entity may be configured to request information about available network function entities providing the service from a Network Repository Function, NRF, entity of the 5G network. The NRF entity may provide the information about available network function entities based on information about network function entities being registered for the service.
[0047] In an exemplary implementation of the network function selection entity, the NWDAF entity is configured to collect the information about the resource required to execute the service and the information about the time required to execute the service from one or more of the network function entities.
[0048] The information about the resource required to execute the service and the time required to execute the service can be flexibly received from one or more of the network function entities. This allows optimal activation of the network functions.
[0049] In an exemplary implementation of the network function selection entity as shown in Figure 3, for example, the network function selection entity is a Service Communication Proxy, SCP, entity of a 5G network which is configured to receive the information about the resource required to execute the service and the information about the time required to execute the service in a service response message from at least one of the network function entities of the 5G network.
[0050] This allows a tradeoff between good task execution by network functions and low resource and time consumptions. For example, in step 2 in Embodiment 2 (see Figure 3), SCP may act as the NFc in Embodiment 1 (see Figure 2), contacting NRF to discover NF1 / NF2 and may contact the NWDAF to collect initial Ri, Ti info provided by NF1 / NF2. Change compared to Embodiment 1 is that the SCP can then monitor updated values provided by the selected NF.
[0051] In an exemplary implementation of the network function selection entity, the SCP entity is configured to monitor whether the received information about the resource required to execute the service and the time required to execute the service meet the threshold criterion.By monitoring the received information, a fast selection of suitable NF entities can be achieved.
[0052] In an exemplary implementation of the network function selection entity, the SCP entity is configured to select at least one of the network function entities, based on updated values of the resource and the time required to execute the service received from at least one of the network function entities in a service response message.
[0053] This allows to adapt to changing network conditions and always to use the actual updated values of resource and time consumption.
[0054] In an exemplary implementation of the network function selection entity as shown in Figure 4, for example, the network function selection entity is a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, entity of a 5G network which is configured to receive the information about the resource and the information about the time in an Nnwdaf_MLModelTraining_Subscribe response message from at least one FL Client NWDAF entity.
[0055] This allows implementation of the current 3GPP standard 29.552 of selection of Vertical Federated Learning clients with inputs resource and time consumption.
[0056] In an exemplary implementation of the network function selection entity, the FL Server NWDAF entity is configured to monitor whether the received information about the resource and the time required to execute the service meet the threshold criterion.
[0057] Monitoring allows to adapt the network function entity selection to changing network conditions.
[0058] In an exemplary implementation of the network function selection entity, the FL server NWDAF entity is configured to select at least one of the FL Client NWDAF entities, based on updated values of the resource and the time received from at least one of the FL Client NWDAF entities in a Nnwdaf_MLModelTraining_Notify message.
[0059] Updating allows to always use the actual, latest values of time and resource consumption when selecting an appropriate network function entity.
[0060] In an exemplary implementation of the network function selection entity as shown in Figure 5, for example, the network function selection entity is a Session Management Function, SMF, entity of a 5G network which is configured to receive the information about the resource and the information about the time required to execute the service in relation to a Policy and Charging Control, PCC, rule in a Policy decision message from a Policy Control Function, PCF entity.
[0061] This allows network function selection under Policy and Charging Control (PCC) rules.
[0062] In an exemplary implementation of the network function selection entity, the SMF entity is configured to monitor whether the received information about the resource and the time required to execute the service in relation to the PCC rule meet the threshold criterion.
[0063] Monitoring allows to adapt the network function entity selection to changing network conditions.
[0064] In an exemplary implementation of the network function selection entity as shown in Figure 6, for example, the processor is configured to select a machine learning algorithm based on the received information about the plurality of network function entities.This allows to implement machine learning and selecting suitable machine learning algorithms.
[0065] In an exemplary implementation of the network function selection entity, the processor is configured to select a Single-task learning, STL, machine learning algorithm or a Multi-task learning, MTL, machine learning algorithm.
[0066] By selecting a suitable machine learning algorithm an optimal policy can be found that maximizes a specific objective, such as long-term rewards.
[0067] In an exemplary implementation of the network function selection entity, the processor is configured to select the machine learning algorithm based on state space information and reward information; wherein the state space information comprises for each service: the service identifier of the service, the received information about the resource and the time required to execute the service and available computing resources of the service; and wherein the reward information comprises a number of completed services considering selected values of the resources and the times of the respective services.
[0068] This allows to select at each step an action from a set of possible actions and transitioning to a new state based on the chosen action. The network function selection entity can find an optimal policy that maximizes a specific objective under the aboveindexed constraints.
[0069] According to a second aspect, the disclosure relates to a network function entity comprising: a processor configured to determine whether the network function entity is capable of providing a service in a communication network, the service being identified by a service identifier; and when the network function entity is capable of providing the service to determine information about a resource required by the network function entity to execute the service and about a time required by the network function entity to execute the service; and a sender configured to transmit the information about the resource and about the time required by the network function entity to execute the service to another network entity.
[0070] Such a network function entity enables provisioning of an effective NF selection in 5G networks enabling the balancing of performance, resource constraints, and user demands to optimize the network's overall functioning. For task execution, two main metrics of 5G Advanced network, which are required Resource (Rf) and required Time (Tf) are exploited. Those parameters can be computed locally by NF entity.
[0071] In an exemplary implementation of the network function entity, the sender is configured to transmit the information about the resource and about the time required by the network function entity to execute the service to a network function selection entity according to the first aspect as described above.
[0072] The information about the resource and about the time required by the network function entity to execute the service can be determined locally be the network function entity which allows a fast selection of an appropriate network function entity by the network function selection entity.
[0073] In an exemplary implementation of the network function entity as shown in Figure 2, for example, the network function entity is a Network Data Analytics Function, NWDAF, entity of the 5G network which is configured to send the information about the resource and time required by a given network function to execute the service to a network function service consumer, NFc, entity of the 5G network.The network function selection is most suitable for the 5G Advanced context, as it considers both 5G Advanced metrics, R and T during the task execution process. The operator is enabled to implement different network function trade-offs, depending on how critical the 5G Advanced task is and depending on the T and R requirements associated with this task.
[0074] In an exemplary implementation of the network function entity as shown in Figure 3, for example, the network function entity is a network function entity of a 5G network which is configured to send the information about the resource and time required to execute the service in a service response message to a network function entity of the 5G network.
[0075] This allows a tradeoff between good task execution by network functions and low resource and time consumptions.
[0076] In an exemplary implementation of the network function entity as shown in Figure 4, for example, the network function entity is a FL Client NWDAF entity of a 5G network which is configured to send the information about the resource and time in an Nnwdaf_MLModelTraining_Subscribe response message to a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, entity of the 5G network.
[0077] This allows implementation of the current 3GPP standard 29.552 of selection of Vertical Federated Learning clients with inputs resource and time consumption.
[0078] In an exemplary implementation of the network function entity as shown in Figure 5, for example, the network function entity is a Policy Control Function, PCF entity of a 5G network which is configured to send the information about the resource and time in relation to a Policy and Charging Control, PCC, rule in a Policy decision message to a Session Management Function, SMF, entity of the 5G network.
[0079] This allows network function selection under Policy and Charging Control (PCC) rules.
[0080] The disclosed network function selection and network function selection entity is most suitable for the 5G Advanced context as compared to the current existing solutions, as it considers both 5G Advanced metrics, R and T during the task execution process. This is a significant advantage of the disclosed solution, as it enables the operator to implement different network functions trade-offs, depending on how critical the 5G Advanced task is, and depending on the T and R requirements associated with this task.
[0081] The presented solution allows optimal activation of the network function based on the trade-offs between accuracy of task execution, event detection, false event detection, R and T.
[0082] The disclosed solution enables to incorporate more parameters to ensure a tradeoff between good task execution by network functions and low resource and time consumptions (generated by the network functions).
[0083] BRIEF DESCRIPTION OF THE DRAWINGS
[0084] Further embodiments of the disclosure will be described with respect to the following figures, in which:
[0085] Figure 1 shows a schematic diagram illustrating a communication network 100 with a network function selection entity 110 according to the disclosure and a plurality of network function entities 121, 122, 123 according to the disclosure;
[0086] Figure 2 shows an exemplary sequence diagram of a 5G network 200 illustrating NF selection based on Ri / Ti;Figure 3 shows an exemplary sequence diagram of a 5G network 300 illustrating delegation of functional selection to SCP:
[0087] Figure 4 shows an exemplary sequence diagram of a 5G network 400 illustrating optimal preparation procedure for Federated Learning;
[0088] Figure 5 shows an exemplary sequence diagram of a 5G network 500 illustrating optimal call flow - Session Establishment; and
[0089] Figure 6 shows an exemplary block diagram of a 5G network 600 illustrating Adaptive ML model selection in 5G NWDAF on dynamic task request scenarios.
[0090] DETAILED DESCRIPTION OF EMBODIMENTS
[0091] In the following detailed description, reference is made to the accompanying drawings, which form a part thereof, and in which is shown by way of illustration specific aspects in which the disclosure may be practiced. It is understood that other aspects may be utilized and structural or logical changes may be made without departing from the scope of the disclosure. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the disclosure is defined by the appended claims.
[0092] It is understood that comments made in connection with a described method may also hold true for a corresponding device or system configured to perform the method and vice versa. For example, if a specific method step is described, a corresponding device may include a unit to perform the described method step, even if such unit is not explicitly described or illustrated in the figures. Further, it is understood that the features of the various exemplary aspects described herein may be combined with each other, unless specifically noted otherwise.
[0093] Figure 1 shows a schematic diagram illustrating a communication network 100 with a network function selection entity 110 according to the disclosure and a plurality of network function entities 121, 122, 123 according to the disclosure.
[0094] The network function selection entity 110 can be used for selecting a network function entity of a plurality of network function entities 121, 122, 123. The network function selection entity 110 comprises: a receiver (not explicitly shown here) configured to collect information 131, 132, 133 about a plurality of network function entities 121, 122, 123 which are capable of providing a service SI (also referred to as a task) in a communication network 100. The service SI is identified by a service identifier 141. The information 131, 132, 133 comprises information about a resource R;i, Ra, Rlnrequired by a respective network function entity of the plurality of network function entities 121, 122, 123 to execute the service SI; and information about a time T;i, Ta, Tlnrequired by the network function entity to execute the service SI.
[0095] The network function selection entity 110 comprises a processor (not explicitly shown here) configured to select 151 at least one network function entity of the plurality of network function entities 121, 122, 123 based on the collected information 131, 132, 133 about resources Rtl, Ra, Rlnand times T;i, Ta, Tlnrequired by the network function entities 121, 122, 123 to execute the service SI which resources and times are matching a threshold criterion.
[0096] The threshold criterion may be based on at least one of the following: load balancing for an evenly load distribution between the network function entities; scalability to scale the network function entity up and down without significant overhead; quality of Service, QoS, to ensure that the selected network function entity can meet QoS requirements of different service types; energy efficiency to reduce energy consumption in the communication network 100.The information about a resource may comprise: a required energy consumption for executing the service SI; a required computation complexity for executing the service S 1; a required communication overhead for executing the service S 1.
[0097] The processor may be configured to select the at least one network function entity 122 which resource Rais below a resource threshold and / or which time Tais below a time threshold.
[0098] The resource threshold may be based on at least one of the following: compute resources availability on one or more edge servers or in a cloud infrastructure; network bandwidth for the service SI; storage capacity; load balancing; CPU usage; memory usage; disk usage; data volume. The time threshold may be based on at least one of the following: latency requirements; response time of the network function entity; processing time of the network function entity; orchestration timing for performing load balancing or scaling actions.
[0099] Figure 1 also shows an exemplary number of network function entities (NF1, NF2, NFn). Each network function entity (e.g., entity 121) comprises: a processor (not shown here) configured to determine whether the network function entity 121 is capable of providing a service SI in a communication network 100. The service SI is identified by a service identifier 141. When the network function entity 121 is capable of providing the service SI, the network function entity 121 is configured to determine information 131 about a resource Rtlrequired by the network function entity 121 to execute the service SI and about a time T;irequired by the network function entity 121 to execute the service SI. The network function entity 121 comprises a sender (not explicitly shown here) configured to transmit the information 131 about the resource Rtland about the time T;irequired by the network function entity 121 to execute the service SI to another network entity.
[0100] The sender may be configured to transmit the information 131 about the resource Rtland about the time T;irequired by the network function entity 121 to execute the service SI to a network function selection entity 110 as described above.
[0101] In particular, Figure 1 illustrates how NF selection is done by NF selection entity 110 based on Rfand Tf.
[0102] 1)- NF selection based on required resources (Rf):
[0103] a)-Compute resources: NFs 121, 122, 123 need specific computational capabilities; resource availability on edge
[0104]
[0105] servers or cloud infrastructure can determine which NF instances can be selected.
[0106] b)-Network Bandwidth: The required bandwidth for a service impacts NF selection; NFs 121, 122, 123 that can't meet bandwidth requirements or are overloaded may be excluded.
[0107] c)-Storage Resources: Some NFs 121, 122, 123 require significant data storage (e.g., for caching content), influencing the selection based on available storage capacity.
[0108] d)-Load Balancing: NF selection aims to distribute workloads evenly across available resources to prevent bottlenecks and overloading certain NFs 121, 122, 123.
[0109] Other parameters related to resource usage may be taken into account.
[0110] 2)-NF selection based on Required Time (Tf):
[0111] a)- Latency requirements: Different applications (e.g., augmented reality, autonomous vehicles) have different latency requirements. NF selection considers the time it takes for a request to be handled, including processing times and network delays. b)-Response Time: For real-time applications, the NF selection process prioritizes instances that can provide a faster response time. This might involve selecting NFs that are geographically closer to the user to reduce propagation delays. ^-Processing Time: The time it takes for an NF 121, 122, 123 to process requests needs to be minimized. NFs that
[0112]
[0113] typically exhibit lower processing times or can be optimized for speed are preferred.d)-Orchestration Timing: Initial load balancing or scaling actions must be performed quickly to prevent service disruptions. An orchestration framework (like Kubemetes) may quickly deploy or relocate NFs based on real-time demand.
[0114] Figure 2 shows an exemplary sequence diagram of a 5G network 200 illustrating NF selection based on Ri / Ti.
[0115] The NF selection entity 210 may correspond to the NF selection entity 110 described above with respect to Figure 1 that has additional features as described in the following.
[0116] The network function selection entity 210 shown in Figure 2 can be a network function service consumer, NFc, of a 5G network 200 which is configured to receive the information 208 about the plurality of network function entities 221,222 from a Network Data Analytics Function, NWDAF, entity 220 of the 5G network 200.
[0117] The NF c entity 210 may be configured to request 207 the information 208 about the plurality of network function entities 221, 222 from the NWDAF entity 220.
[0118] The NWDAF entity 220 may be configured to collect 203, 204 the information about the resource Rtand the information about the time T;required to execute service SI from one or more of the network function entities (NFsl, NF s2) 221, 222.
[0119] The network function entity shown in Figure 2 can be a Network Data Analytics Function, NWDAF, entity 220 of the 5G network 200 which is configured to send the information 208 about the resource Rtland time T;ito a network function service consumer, NFc, entity 210 of the 5G network 200.
[0120] In particular, in the 5G network 200 shown in Figure 2, the NWDAF (eNWDAF) is enhanced to monitor required resource Rtand required T;per NF and per service. This functionality can be added to NWDAF but can also be supported by another NF, e.g. NRF.
[0121] NF service producers NF s 1 and NF s2 can be two interchangeable NF service instances of the same service type S 1.
[0122] The following sequence as shown in Figure 2 can be performed by those entities:
[0123] 1. (Registration 201 ofNFsl for Service SI):
[0124] NFsl, 221 registers 201 to NRF 230 to inform the NRF of its NF profile when the NF service instance becomes operative for the first time. The NRF 230 stores the NF profile of NF service instance NFsl, 221 and marks the NF service instance available. As an option, required Ri and Ti for service SI can be included in the NF profile stored in the NRF 230.
[0125] 2. (Registration 202 ofNFs2 for Service SI):
[0126] NFs2, 222 registers to NRF 230 to inform the NRF of its NF profile when the NF service instance becomes operative for the first time. The NRF 230 stores the NF profile of NF service instance NFs2, 222 and marks the NF service instance available. As an option, required Rtand T;for service SI can be included in the NF profile stored in the NRF 230.3. (Collection 203 of Ri, Ti for service SI):
[0127] The Data Collection from NFsl, 221 is used by NWDAF 220 to subscribe / unsubscribe at NFs1 221 to be notified for the required Riand required Tifor the service SI. NWDAF 220 can discover NFsl, 221 via NRF 230 or be directly contacted by NFsl, 221. NF service NFsl regularly updates the NWDAF 220 with current R. andT;.
[0128] 4. (Collection 204 of Ri, Ti for service SI):
[0129] The Data Collection from NF s2, 222 is used by NWDAF 220 to subscribe / unsubscribe at NFs2, 222 to be notified for the required Riand required Tifor the service SI. NWDAF 220 can discover NFs2, 222 via NRF 230 or be directly contacted by NFs2, 222. NF service NF s2 regularly updates the NWDAF 220 with current Ri and Ti.
[0130] 5. (NFs for Service SI, 205):
[0131] The NF service consumer NFcl, 210 queries the NRF 230 to discover NF service instances available for service SI in the network based on service name.
[0132] 6. (NFs1 and NFs2 available for Service SI, 206):
[0133] The NRF 230 determines a set of NF instances (NFsl, 221 and NFs2, 222) matching the discovery request sent by NFcl, 210 and sends the NF profile(s) of NFs1 and NFs2 instances. As an option, required Riand Tifor service SI can be included in the NF profile sent by the NRF 230.
[0134] 7. (Ri and Ti for Service SI for NFsl and NFs2, 207):
[0135] In order to select the most suitable service instance, NFcl, 210 contacts the NWDAF 220 to retrieve the Riand Tirequired by NFs1, 221 and NFs2, 222 for service S1.
[0136] 8. (Latest Riand Tifor Service SI for NFs1 and NFs2, 208):
[0137] NWDAF 220 provides Riand Ti.
[0138] 9. (Service request sent to NFs2, 209):
[0139] NFcl, 210 sends service request to NFs2, 222.
[0140] Figure 3 shows an exemplary sequence diagram of a 5G network 300 illustrating delegation of functional selection to SCP.
[0141] The NF selection entity 310 may correspond to the NF selection entity 110 described above with respect to Figure 1 that has additional features as described in the following.
[0142] The NF selection entity 310 shown in Figure 3 can be a session control plane, SCP, of a 5G network 300 which is configured to receive the information about the resource Riand the time Tiin a service response message 304 from at least one of the network function entities 321, 322 of the 5G network 300.
[0143] The SCP entity 310 may be configured to monitor whether the received information about the resource Riand the time Timeet the threshold criterion.
[0144] The SCP entity 310 may be configured to select at least one of the network function entities 321, 322, based on updated values of the resource Riand the time Tireceived from at least one of the network function entities 321, 322 in a service response message 304.The network function entity can be a network function entity 321 of a 5G network 300 as shown in Figure 3 which is configured to send the information 131 about the resource Riand time Tiin a service response message 304 to a network function selection entity 310 of the 5G network 300.
[0145] The following sequence as shown in Figure 3 can be performed by those entities:
[0146] The Session Control Plane (SCP) 310 performs NF selection based on required resources and time by following a systematic approach.
[0147] 1)-Resource Requirements:
[0148] a)-SCP 310 queries the Network Repository Function (NRF) to retrieve information about available NFs.
[0149] b)-SCP filters the NF s based on their resource capabilities, such as processing power, memory, and bandwidth, to match the required resources for the service.
[0150] 2)-Time requirements:
[0151] a)-SCP 310 considers the latency requirements of the service. NFs that can meet the required latency are prioritized. b)-SCP ensures that the selected NFs comply with the service guaranteeing timely delivery of resources.
[0152] SCP continuously monitors network conditions and dynamically adjusts NF selection to maintain optimal performance. SCP uses predictive analytics to anticipate network load and preemptively select NFs that can handle the expected demand.
[0153] Figure 3 illustrates NFps selection taken as inputs the parameters R, and Tj. Three iterations i (300a), i+1 (300b) and i+2 (300c) are illustrated in Figure 3.
[0154] In Step 1, the NFcl, 330 sends a service request 301 with parameters target-nf-set-id=Setl to SCP 310.
[0155] In Steps 2 to 5 (302, 303, 304, 305, 306) of first iteration 300a, the SCP 310 performs selection among NFs belonging to the received target set at the iteration t.
[0156] In Steps 6 to 10 (307, 308, 309, 310b, 311) of second iteration 300b, the SCP 310 performs selection among NFs belonging to the received target set at the iteration t+1.
[0157] In Steps 2, 7, 13 (303, 308, 314), when R, and T, are greater than predefined thresholds (those thresholds are left to the implementation, e.g., using the defined functions), SCP 310 requests 303, 308, 314 theNFpi, 321, 322 to take into account the resource constraints and computation time (i.e., decrease R, and T.
[0158] In Steps 5, 10 and 10 (306, 311, 317), the selection of suitable NFpi is based on the last updated values of parameters, Ri;and Tj. i.e., RJt+l) « RJt) and TJt+l) « T(t).
[0159] Figure 4 shows an exemplary sequence diagram of a 5G network 400 illustrating optimal preparation procedure for Federated Learning.
[0160] The NF selection entity 410 may correspond to the NF selection entity 110 described above with respect to Figure 1 that has additional features as described in the following.The NF selection entity 410 shown in Figure 4 can be a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, of a 5G network 400 which is configured to receive the information about the resource Riand the information about the time Tiin an Nnwdaf_MLModelTraining_Subscribe response message 404 from at least one FL Client NWDAF entity 420.
[0161] The FL Server NWDAF entity 410 may be configured to monitor 405 whether the received information about the resource Riand the time Timeet the threshold criterion.
[0162] The FL server NWDAF entity 410 may be configured to select 407 at least one of the FL Client NWDAF entities 420, based on updated values of the resource Riand the time Tireceived from at least one of the FL Client NWDAF entities 420 in a Nnwdaf_MLModelTraining_Notify message 406.
[0163] The network function entity can be a FL Client NWDAF entity 420 of a 5G network 400 which is configured to send the information 131 about the resource Ri and time Ti in an Nnwdaf_MLModelTraining_Subscribe response message 404 to a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, entity 410 of the 5G network 400.
[0164] Figure 4 illustrates an embodiment example on how to improve the current 3GPP Standard 29.552 of selection of Vertical Federated Learning (VFL) client taken the inputs Riand Ti: Two iterations i (400a) and i+1 (400b) are illustrated in Figure 4.
[0165] 1. The FL Server NWDAF 410 and FL Client NWDAF(s) 420 are discovered 401 via NRF 430. Details are described in clause 5.3.2.2 of 3GPP TS 29.510.
[0166] 2. The FL Server NWDAF 410 invokes Nnwdaf_MLModelTraining_Subscribe service operation 402 by sending an HTTP POST request targeting the resource " NWDAF ML Model Training Subscriptions", The request shall include the "mLPreFlag" attribute and set to "true".
[0167] 3. The FL Client NWDAF(s) 420 decides 403 whether to join the FL process based on implementation.
[0168] 4. The FL Client NWDAF 420 responds to the Nnwdaf_MLModelTraining_Subscribe request 402 to indicate its decision. This request includes also the parameters, Riand Ti.
[0169] 5. The FL Server NWDAF 410 monitors 405 the parameters, Riand Tiby using the Nnwdaf_MLModelMonitor service, and stores the values of Riand Ti.
[0170] 6. The FL Client NWDAF 420 periodically reports 406 (using Notify) to the FL Server NWDAF 410 updated Riand Ti. In the present case, the Riand Tiare greater than predefined thresholds (those thresholds are left to the implementation, e.g., using the defined functions).
[0171] 7. If Ri≥ Tiand / or If Ti≥ Ti; the FL Server NWDAF 410 decides to select 407 another FL Client NWDAF for the FL process.
[0172] Figure 5 shows an exemplary sequence diagram of a 5G network 500 illustrating optimal call flow - Session Establishment.
[0173] The NF selection entity 510 may correspond to the NF selection entity 110 described above with respect to Figure 1 that has additional features as described in the following.The NF selection entity 510 shown in Figure 5 can be a Session Management Function, SMF, of a 5G network 500 which is configured to receive the information about the resource Riand the information about the time Tiin relation to a Policy and Charging Control, PCC, rule in a Policy decision message 506 from a Policy Control Function, PCF entity 520.
[0174] The SMF entity 510 may be configured to monitor 507 whether the received information about the resource Riand the time Tiin relation to the PCC rule meet the threshold criterion.
[0175] The network function entity can be a Policy Control Function, PCF entity 520 of a 5G network 500 which is configured to send the information 506 about the resource Ri, and time Tiin relation to a Policy and Charging Control, PCC, rule in a Policy decision message 506 to a Session Management Function, SMF, entity 510 of the 5G network 500.
[0176] In particular, Figure 5 illustrates how SMF 510 selects 504 PCF 520 at the iteration t (500a) considering the optimal values of R and T. At the iteration t+1 (500b) the following processes are executed between the SMF 510 and PCF 520:
[0177] 5. The SMF 510 sends SM policy association establishment request 505 to PCF 520. In this message the SMF requests the PCF, the maximum values of of Riand Tirequired on executing the specific rules.
[0178] 6. The PCF 520 replies by sending a Policy decision message 506. This message includes the required Riand Tirelated to each PCC rule.
[0179] 7. SMF 510 monitors 507 the values of Riand Tirelated to each PCC rule. In case, when Riand Tiare greater than predefined threshold (thresholds are left to the implementation, e.g., using the defined functions), those related PCC rules are not selected. Therefore, only the PCC rules that require low Riand Tiare selected.
[0180] 8. SMF 510 has further exchanges with the PCF 520.
[0181] 9. Response 509 from the PCF 520 includes updated Ri, Ti.
[0182] 10. When updated Ri, Tiare not suitable, SMF 510 selects 510b another PCF.
[0183] Figure 6 shows an exemplary block diagram of a 5G network 600 illustrating Adaptive ML model selection in 5G NWDAF on dynamic task request scenarios.
[0184] The NF selection entity 610 may correspond to the NF selection entity 110 described above with respect to Figure 1 that has additional features as described in the following.
[0185] The NF selection entity 610 shown in Figure 6 can be an agent, e.g., a NWDAF.
[0186] The processor of the NF selection entity 610 may be configured to select 601 a machine learning algorithm 611, 612 based on the received information about the plurality of network function entities.
[0187] The processor of the NF selection entity 610 may be configured to select 601 a Single-task learning, STL, machine learning algorithm 611 or a Multi-task learning, MTL, machine learning algorithm 612.The processor of the NF selection entity 610 may be configured to select 601 the machine learning algorithm 611, 612 based on state space information 602 and reward information 603. The state space information 602 may comprise for each service: the service identifier of the service, the received information about the resources Ri and the times Ti of the service and available computing resources of the service. The reward information 603 may comprise a number of completed services considering selected values of the resources Ri and the times Ti of the respective services.
[0188] Machine learning algorithm selection may be initiated when NF’s task request arrives at NWDAF 610. A requested task may include the required latency and various types of analytics, such as user mobility prediction and slice load prediction. As shown in Figure 6, NWDAF 610 can leverage this information, along with Riand Tirelated to each requested task to make decisions in selecting Machine learning algorithm (STL or MTL) on dynamic task request scenarios. Note that Single-task learning is STL and Multi-task learning is MTL.
[0189] NWDAF 610 can be an agent in this system that interacts with an environment:
[0190] At each step, an action may be selected from a set of possible actions, and the environment responds by transitioning to a new state based on the chosen action. The goal is to find an optimal policy that maximizes a specific objective, such as long-term rewards. The agent can be modeled as follows:
[0191] • State space S = {available computing resource, task i, Riand Tiof each task}
[0192] • Action space A = {Single-task learning (STL), Multi-task learning (MTL)}
[0193] • Reward R = (number of completed tasks considered the optimal values of Riand Ti)
[0194] While a particular feature or aspect of the disclosure may have been disclosed with respect to only one of several implementations, such feature or aspect may be combined with one or more other features or aspects of the other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms "include", "have", "with", or other variants thereof are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term "comprise". Also, the terms "exemplary", "for example" and "e.g." are merely meant as an example, rather than the best or optimal. The terms “coupled” and “connected”, along with derivatives may have been used. It should be understood that these terms may have been used to indicate that two elements cooperate or interact with each other regardless whether they are in direct physical or electrical contact, or they are not in direct contact with each other.
[0195] Although specific aspects have been illustrated and described herein, it will be appreciated by those of ordinary skill in the art that a variety of alternate and / or equivalent implementations may be substituted for the specific aspects shown and described without departing from the scope of the disclosure. This application is intended to cover any adaptations or variations of the specific aspects discussed herein.
[0196] Although the elements in the following claims are recited in a particular sequence with corresponding labeling, unless the claim recitations otherwise imply a particular sequence for implementing some or all of those elements, those elements are not necessarily intended to be limited to being implemented in that particular sequence.
[0197] Many alternatives, modifications, and variations will be apparent to those skilled in the art in light of the above teachings. Of course, those skilled in the art readily recognize that there are numerous applications of the disclosure beyond those described herein. While the disclosure has been described with reference to one or more particular embodiments, those skilled in the art recognize that many changes may be made thereto without departing from the scope of the disclosure. It is therefore to beunderstood that within the scope of the appended claims and their equivalents, the disclosure may be practiced otherwise than as specifically described herein.
Claims
CLAIMS1. A network function selection entity (110) for selecting a network function entity of a plurality of network function entities (121, 122, 123), the network function selection entity (110) comprising:a receiver configured to collect information (131, 132, 133) about a plurality of network function entities (121, 122, 123) which are capable of providing a service (SI) in a communication network (100), the service (SI) being identified by a service identifier (141);wherein the information (131, 132, 133) comprises information about a resource (Ril, Ri2, Rin) required by a respective network function entity of the plurality of network function entities (121, 122, 123) to execute the service (SI); and information about a time (Til, Ti2, Tin) required by the network function entity to execute the service (SI); and a processor configured to select (151) at least one network function entity of the plurality of network function entities (121, 122, 123) based on the collected information (131, 132, 133)about resources (Ril, Ri2, Rin) and times (Til, Ti2, Tin) by the network function entities (121, 122, 123) required to execute the service (SI) which resources and times are matching a threshold criterion.
2. The network function selection entity (110) of claim 1, wherein the threshold criterion is based on at least one of the following:load balancing for an evenly load distribution between the network function entities;scalability to scale the network function entity up and down without significant overhead;quality of Service, QoS, to ensure that the selected network function entity can meet QoS requirements of different service types;energy efficiency to reduce energy consumption in the communication network (100).
3. The network function selection entity (110) of claim 1 or 2, wherein the information about a resource comprises: a required energy consumption for executing the service (SI);a required computation complexity for executing the service (SI);a required communication overhead for executing the service (SI).
4. The network function selection entity (110) of any of the preceding claims,wherein the processor is configured to select the at least one network function entity (122) which resource (Ri2) is below a resource threshold and / or which time (Ti2) is below a time threshold.
5. The network function selection entity (110) of any of the preceding claims,wherein the resource threshold is based on at least one of the following:compute resources availability on one or more edge servers or in a cloud infrastructure;network bandwidth for the service (SI);storage capacity;load balancing;CPU usage;memory usage;disk usage;data volume; andwherein the time threshold is based on at least one of the following:latency requirements;response time of the network function entity;processing time of the network function entity;orchestration timing for performing load balancing or scaling actions.
6. The network function selection entity (110) of any of the preceding claims,being a network function service consumer, NFc, entity (210) of a 5G network (200) which is configured to receive the information (208) about the plurality of network function entities (221, 222) from a Network Data Analytics Function, NWDAF, entity (220) of the 5G network (200).
7. The network function selection entity (110) of claim 6,wherein the NFc entity (210) is configured to request (207) the information (208) about the plurality of network function entities (221, 222) from the NWDAF entity (220).
8. The network function selection entity (110) of claim 6 or 7,wherein the NWDAF entity (220) is configured to collect (203, 204) the information about the resource (Ri) and the information about the time (Ti) required to execute the service (SI) from one or more of the network function entities (221, 222).
9. The network function selection entity (110) of claims 1 to 5,being a session control plane, SCP, entity (310) of a 5G network (300) which is configured to receive the information about the resource (Ri) and the information about the time (Ti) required to execute the service (SI) in a service response message (304) from at least one of the network function entities (321, 322) of the 5G network (300).
10. The network function selection entity (110) of claim 9,wherein the SCP entity (310) is configured to monitor whether the received information about the resource (Ri) and the time (Ti) meet the threshold criterion.
11. The network function selection entity (110) of claim 10,wherein the SCP entity (310) is configured to select (306) at least one of the network function entities (321, 322), based on updated values of the resource (Ri) and the time (Ti) received from at least one of the network function entities (321, 322) in a service response message (305).
12. The network function selection entity (110) of claims 1 to 5,being a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, entity (410) of a 5G network (400) which is configured to receive the information about the resource (Ri) and the information about the time (Ti) in an Nnwdaf_MLModelTraining_Subscribe response message (404) from at least one FL Client NWDAF entity (420).
13. The network function selection entity (110) of claim 12,wherein the FL Server NWDAF entity (410) is configured to monitor (405) whether the received information about the resource (Ri) and the time (Ti) required to execute the service (SI) meet the threshold criterion.
14. The network function selection entity ( 110) of claim 13,wherein the FL server NWDAF entity (410) is configured to select (407) at least one of the FL Client NWDAF entities (420), based on updated values of the resource (Ri) and the time (Ti) received from at least one of the FL Client NWDAF entities (420) in a Nnwdaf_MLModelTraining_Notify message (406).
15. The network function selection entity (110) of claims 1 to 5,being a Session Management Function, SMF, entity (510) of a 5G network (500) which is configured to receive the information about the resource (Ri) and the information about the time (Ti) in relation to a Policy and Charging Control, PCC, rule in a Policy decision message (506) from a Policy Control Function, PCF entity (520).
16. The network function selection entity ( 110) of claim 15,wherein the SMF entity (510) is configured to monitor (507) whether the received information about the resource (Ri) and the time (Ti) in relation to the PCC rule meet the threshold criterion.
17. The network function selection entity (110) of any of the preceding claims,wherein the processor is configured to select (601) a machine learning algorithm (611, 612) based on the received information about the plurality of network function entities.
18. The network function selection entity ( 110) of claim 17,wherein the processor is configured to select (601) a Single-task learning, STL, machine learning algorithm (611) or a Multi-task learning, MTL, machine learning algorithm (612).
19. The network function selection entity (110) of claim 17 or 18,wherein the processor is configured to select (601) the machine learning algorithm (611, 612) based on state space information (602) and reward information (603);wherein the state space information (602) comprises for each service: the service identifier of the service, the received information about the resources (Ri) and the times (Ti) required to execute the service and available computing resources of the service; andwherein the reward information (603) comprises a number of completed services considering selected values of the resources (Ri) and the times (Ti) required to execute the respective services.
20. A network function entity (121) comprising:a processor configured to determine whether the network function entity (121 ) is capable of providing a service (SI) in a communication network (100), the service (SI) being identified by a service identifier (141); and when the network function entity (121) is capable of providing the service (SI) to determine information (131) about a resource (Ril) required by the network function entity (121) to execute the service (SI) and about a time (Til) required by the network function entity (121) to execute the service (SI); anda sender configured to transmit the information (131) about the resource (Ril ) and about the time (Til ) required by the network function entity (121) to execute the service (SI) to another network entity.
21. The network function entity (121) of claim 20,wherein the sender is configured to transmit the information (131) about the resource (Ril) and about the time (Til) required by the network function entity (121) to execute the service (SI) to a network function selection entity (110) according to any of claims 1 to 19.
22. The network function entity (121) of claim 20 or 21,being a Network Data Analytics Function, NWDAF, entity (220) of the 5G network (200) which is configured to send the information (208) about the resource (Ril) and time (Til) to a network function service consumer, NFc, entity (210) of the 5G network (200).
23. The network function entity (121) of claim 20 or 21,being a network function entity (321) of a 5G network (300) which is configured to send the information (131) about the resource (Ri) and time (Ti) required to execute the service (SI) in a service response message (304) to a network function entity (321) of the 5G network (300).
24. The network function entity (121) of claim 20 or 21,being a FL Client NWDAF entity (420) of a 5G network (400) which is configured to send the information (131) about the resource (Ri) and time (Ti) in an Nnwdaf_MLModelTraining_Subscribe response message (404) to a Federated Learning, FL, Server Network Data Analytics Function, NWDAF, entity (410) of the 5G network (400).
5. The network function entity ( 121 ) of claim 20 or 21,being a Policy Control Function, PCF entity (520) of a 5G network (500) which is configured to send the information (506) about the resource (Ri) and time (Ti) in relation to a Policy and Charging Control, PCC, rule in a Policy decision message (506) to a Session Management Function, SMF, entity (510) of the 5G network (500).