Admission Control Method and Admission Request Method in a Communication Network System

Abstract

In a communication network system that deploys a plurality of network slices, a slice admission control method executed by a slice controller includes: a) receiving a slice request including slice request parameters from a user equipment; b) obtaining network parameters related to at least an overall state of network resources in the plurality of network slices; c) for each network slice, a slice service variable related to the network resources required from the network slice for permitting the slice request, which is obtained based on the slice request parameters and the network parameters, a slice utility gain expected from the utilization of the network resources required from the network slice for permitting the slice request, which is obtained based on a machine learning model, and a slice demand load related to the network resources consumed in the network slice when the slice request is permitted, which is obtained based on the slice utility gain of the network slice; d) transmitting the slice demand load to the user equipment; e) supplying the slice demand load to the machine learning model; and f) determining a decision regarding the admission of the slice request based on the slice utility gain.

Description

Technical Field

[0001] The present disclosure relates to the field of resource management in network slicing, and more particularly to the optimization of slice admission control mechanisms in sliced 5G networks.

Background Art

[0002] The introduction of the fifth-generation mobile network (5G) for wireless communication aims to enhance service flexibility while considering latency, reliability, and / or other quality of service (QoS) requirements of users. In this introduction, it is necessary to accommodate the diversity of various 5G service types. In such various 5G service types, service requirements and resource requirements may differ with respect to throughput, access capacity, and latency. To accommodate such diversity, network slicing is configured by managing multiple independent logical networks (i.e., multiple network slices) on a shared physical network infrastructure. Each network slice is flexible and scalable by using flexible virtualized network function (NF) instances, and supports, for example, specific use cases with various requirements for speed, reliability, throughput, and / or energy efficiency. Therefore, resources including, for example, core resources, radio resources, and transfer resources are allocated to each network slice to meet incoming user requests.

[0003] The network capacity requirements of a network slice are managed by the application layer of the network. This layer is composed of an edge computing platform. The application layer manages the slice level specification (SLS) associated with the network slice that includes the required service quality parameters. Such slice level specifications are a set of service level requirements associated with the service level agreement (SLA) to be met by the network slice. Such specifications may vary for each network slice and can determine the service profiles of user equipment (UE) using different network slices.

[0004] The main challenge in the management of network slices is to optimize the allocation of network slices to incoming user requests while meeting the respective requirements of each slice. This is called the slice admission control policy. For example, a user may request a specific throughput on a specific network slice. However, the total throughput allocated to all users of such a specific network slice is restricted (thus limited) by the resources allocated to such a network slice in the application layer. Such a slice admission control policy can be executed, for example, by a slice controller (or slice orchestrator) of the network infrastructure.

[0005] The slice admission control policy consists of receiving and processing user slice requests in order to utilize network slices for the execution of specific services and applications according to specific service requirements. Each user slice request may refer to a user request to transmit a specific throughput across a network slice. The slice admission control policy aims to maximize network resource usage by accepting (or permitting) as many user slice requests as possible while considering the necessary slice-level specifications and the limited amount of physical resources of the network allocated to the network slice. Therefore, it is particularly complex to determine an optimized slice admission control policy. Thus, the slice admission control policy consists of permitting or rejecting incoming user slice requests while optimizing the use of network capabilities.

[0006] Existing methods of slice admission control mechanisms are based on the premise that user slice requests are received by the slice controller in an unregulated and random manner. Also, such existing methods assume that service requests from users regarding service type or quality of service requirements are random and cannot be negotiated or predicted. As a result, the network infrastructure is exposed to an unpredictable situation where the network resources of the network slice are insufficient or not used according to the timing and requirements of incoming user slice requests. In particular, there is a risk that collisions may occur regularly between multiple incoming requests for the same network slice, and these collisions may cause the slice controller to reject the admission of many slice requests due to the slice malfunction. When there are multiple incoming requests for the same network slice and the slice controller cannot process them simultaneously, the slice controller may also buffer slice requests for a time that is unacceptable with respect to the delay tolerance of the service requesting slice resources.

[0007] In addition, in existing methods for managing slice admission control, the requirements of slice-level specifications determined by the application layer are considered fixed. As a result, existing slice admission control algorithms consist of determining whether to permit or reject slice requests based on fixed constraints regarding application requirements. However, many, especially less critical 5G applications, can have flexible requirements. Applications with such flexible specifications can adjust their requirements using the resources available in the network slice. In existing methods, such application flexibility is not considered in the slice admission control policy.

[0008] Existing slice admission control algorithms also face the challenges of complexity and low convergence time due to the large number of network slices and the variety of network states of the communication network that need to be considered when receiving slice requests.

Summary of the Invention

Problems to be Solved by the Invention

[0009] The present disclosure aims to improve the above situation and optimize the network slice admission control policy.

Means for Solving the Problems

[0010] Therefore, an admission control method executed by a slice controller for performing slice admission control in a communication network system, where the communication network system deploys a plurality of network slices, The admission control method includes: a) receiving at least one slice admission request from at least one user equipment deployed in the communication network system, the slice admission request including data related to at least slice request parameters; b) obtaining network parameters related to at least the overall state of network resources in a plurality of network slices; c) for each network slice of the plurality of network slices, a slice service variable related to the network resources required from the network slice to permit a slice admission request, the slice service variable being determined based on at least slice request parameters and network parameters; a slice utility gain related to a change in a slice utility function expected from the utilization of the network resources required from the network slice to permit a slice admission request, the slice utility function being dependent on slice service variables respectively associated with the plurality of network slices, the slice utility gain being determined based on at least a machine learning model; when the slice admission request is permitted, a slice demand load related to the network resources consumed in the network slice, the slice demand load being determined based on at least the slice utility gain of the network slice; determining; d) transmitting the slice demand loads respectively associated with the plurality of network slices to a user equipment; e) supplying the slice demand loads respectively associated with the plurality of network slices to a machine learning model; f) making a decision regarding permission of the slice admission request based on the slice utility gain; and proposing an admission control method including the above.

[0011] As a result, with the proposed admission control method, a communication network system can manage the allocation of network resources in a plurality of network slices for a plurality of incoming slice admission requests. In particular, with the proposed admission control method, the allocation of network resources in a network slice can be optimized by considering both the required parameters derived from an incoming slice admission request and the potential impact of such a request on the network resources of the system.

[0012] Also, the proposed admission control method depends on the slice utility gain obtained based on a machine learning model considering the slice demand load. As a result, the admission control policy is executed considering the slice utility gains of all network slices (e.g., by maximizing such slice utility gains). Such slice utility gains depend on the slice demand load resulting from the slice admission requests executed by user equipment. In particular, such slice demand load is fed to the same machine learning model that enables the estimation and / or prediction of the slice utility gain of each network slice and is learned by that machine learning model.

[0013] Therefore, the proposed admission control method uses a unique machine learning model to directly associate the service utility gains of multiple network slices with the impact of potentially admitted slice admission requests via slice demand load. The proposed admission control method enables the formation of slice admission requests by sending such slice demand load to user equipment. In fact, the proposed admission control method proposes to provide user equipment with information related to the resource load on each network slice due to incoming slice admission requests by sending slice demand load to the user equipment. Such resource load indicates that some slice admission requests may be rejected due to network resource outages in some network slices. On the other hand, based on the slice demand load sent in this way, both the user equipment (i.e., the side requesting resources) and the slice controller (i.e., the side allocating resources) recognize the resource capabilities of all network slices. In particular, the user equipment can adapt its request method according to such slice demand load. For example, based on such slice demand load, further slice admission requests can be directed to the network slice with the lowest slice demand load. In other words, the proposed admission control method can adjust the slice requests sent by the user equipment within the admission control mechanism. In particular, the proposed admission control method introduces slice demand load to challenge the unpredictable and unregulated random nature of incoming user admission requests, which is assumed in existing slice admission control mechanisms. Such slice demand load is fed into and learned by a machine learning model that enables prediction and / or estimation of the slice utility gain of network slices. For example, in the case of a machine learning model with a neural network model, a unique neural network can be implemented to learn about both the slice utility gain and the slice demand load of network slices.

[0014] Furthermore, when the slice demand load is dynamically updated, at least part of the content of the slice admission request becomes predictable. In fact, since the incoming slice admission request is related to the network slice with the lowest slice demand load, the computational complexity and convergence time of the machine learning model for executing the slice admission control method may be reduced in the long term.

[0015] The slice controller can be understood as a computing unit of a communication network system that executes a slice management control mechanism within the network system. In other words, the slice controller processes the incoming slice admission requests from user equipment, centrally manages information within the network system, and determines the permission of such slice admission requests, that is, it can determine the permission or rejection of slice admission requests. Therefore, the slice controller can determine the allocation of network resources in a plurality of network slices deployed by the network system. In particular, the slice controller can optimize the allocation of network resources for all network slices by depending on a machine learning model when executing the slice admission control mechanism.

[0016] A communication network system can be understood as an aggregate of computer components that includes hardware modules and software modules that support services and applications of users (user devices) via a communication network. The communication network can be a software-defined network system that includes an infrastructure layer, a control layer, and an application layer. The communication network system can be, for example, a fifth-generation (5G) system. The communication network system supports the communication of user devices by allocating network resources (e.g., from a resource pool) to such user devices in order to exchange data signals and / or control signals via, for example, the communication channels of the network system. In particular, the communication network system can deploy a plurality of virtually separated networks, also called network slices, across a common physical infrastructure in order to meet a plurality of service level specifications (SLS) and some service quality requirements.

[0017] A network slice can be understood as a virtual network that uses a part of the radio access network of a communication network. The network slice deploys network functions that guarantee the transfer of slice-related packets having specific quality of service (QoS) requirements. Network resources can be allocated to each network slice in order to provide applications and services for a plurality of user devices. The network resources allocated to such a network slice can be set or configured, for example, by the application layer of the communication network system.

[0018] A user device can be understood as any electronic device capable of communicating via a communication network system. Such user devices can include, for example, computer devices, connected devices, connected vehicles, mobile devices such as mobile phones, laptops, tablets, etc. A user device may also refer to any electronic device (IoT (Internet of Things) device) from the Internet of Things.

[0019] A user device deployed in a communication network system can be understood as a user device that can receive resource allocation from the communication network system, for example, via a base station and / or within a cell. The user device can connect to the communication network system, for example, after performing a registration process or an initial access procedure to the network system in advance. The user device physically belongs to the deployment area of the communication network system and can request network resources of the network system.

[0020] Slice admission control can be understood as an admission control mechanism executed by a slice controller of a communication network system to allocate network resources to an incoming slice admission request sent by one or several of the user devices. In particular, such slice admission control can adjust the allocation of network resources in the long term to optimize the service utility gain of the network system. In other words, slice admission control is a slice admission request processing mechanism for maximizing the use of network resources provided by the network system (i.e., maximizing by permitting as many slice requests as possible) while guaranteeing the quality of service and service performance required by the user device. Slice admission control can rely on a machine learning model to optimize the service utility gain of the network system in the long term.

[0021] A slice admission request (also referred to as a slice admission demand, slice request, or slice demand) can be understood as a demand (or request) for traffic (e.g., throughput demand) in at least one network slice of a communication network, and thus a demand for resource allocation. Such a demand is sent by a user equipment. Such a demand can refer to, for example, a request to transmit data packets (or application-level packets) using resources allocated to at least one specific network slice of a communication network. The demand can include specific requirements regarding quality of service parameters such as delay parameters, reliability parameters, and / or throughput parameters. For example, a slice request can include a request for a minimum value of a throughput parameter and / or a maximum value of a delay (i.e., delay tolerance level) for the transmission of application-level packets.

[0022] Data related to slice request parameters can be understood as parameters that characterize each slice request received by a slice controller. In particular, slice request parameters can define the service requirements of a user equipment in a slice request. For example, slice request parameters can include the throughput required in one or more network slices. Slice request parameters can also include, for example, data related to a delay tolerance level.

[0023] The network parameters related to the overall state of network resources in multiple network slices can be understood as information specifying the nature and / or amount of available network resources allocated to the network slices. In other words, such network parameters can be understood as reflecting the state of the network system regarding the available resources. Such network parameters can include, for example, the current amounts of available cloud resources, computing resources, storage resources, and memory resources. Such network parameters can be provided, for example, by subsystems of the communication network system. Such subsystems include, for example, the core network of the communication network system, edge computing, shared access network, and / or transfer network.

[0024] Obtaining the parameters (e.g., slice service variable, slice utility gain, slice utility function, slice demand load) of each network slice in multiple network slices can be understood as the parameters being reasonable for each network slice of the communication network system, considering the current state of the network system at the time when such parameters are obtained.

[0025] A slice service variable is a parameter associated with each network slice of a network system and can be understood as the overall demand parameter requested by a user across the network slices. For example, the slice service variable can be understood as the overall throughput requested by multiple incoming slice requests across the corresponding network slice. In other words, the slice service variable can refer to the sum of the demand parameters (e.g., throughput) of user equipment accessing the network slice, which are requested in each slice request. In another embodiment, the slice service variable can be understood as the sum of the throughputs respectively requested via slice requests by user equipment that requests access to the corresponding network slice. In yet another embodiment, the service slice variable can be understood as the average of the demand parameters of user terminals accessing the network slice, or the average demand parameter of user terminals requesting access to the network slice. More generally, the slice service variable can be understood as a parameter that reflects the updated state and service quality parameters (throughput, reliability, security, latency, etc.) that can be provided by such a network slice when the corresponding network slice distributes demand parameters to incoming user requests that demand such parameters for that network slice.

[0026] The slice utility gain can be understood as the change in the utility function associated with the use of network resources of a network slice. Therefore, a positive slice utility gain can indicate that the operation under consideration (e.g., the operation of permitting a slice request in a network slice) generates a positive gain in the utility of the corresponding network slice. On the other hand, a negative slice utility gain can indicate that the operation under consideration generates a negative gain (i.e., a loss) in the utility of the corresponding network slice. For example, if a slice request is permitted even though the network resources allocated to such a network slice are insufficient to satisfy both the demand parameters of the slice request and the requirements of other user equipment already permitted in the network slice, the slice utility gain may be negative. The slice utility gain can be understood, for example, as the derivative of the slice utility function associated with the network slice with respect to the corresponding slice service variable.

[0027] The slice utility function can be understood as a multivariable function associated with a network slice and reflecting the utility in the use of network resources of that network slice. In particular, such a slice utility function can be dependent on the slice service variables of all network slices deployed by the network system.

[0028] The slice demand load can be understood as a parameter reflecting the resource cost of permitting an incoming slice request for the network slice under consideration. Therefore, the slice demand load can be understood as the "resource price" at which the network slice permits an incoming slice request. Thus, the greater the slice demand load of the network slice under consideration, the greater the resource consumption of the user request.

[0029] A machine learning model can be understood as a learning algorithm such as a reinforcement learning algorithm, a neural network algorithm, a deep neural network algorithm, and / or a deep reinforcement learning model. In particular, the machine learning model maps a slice admission control mechanism that provides the operations performed by the slice controller based on at least a set of network states, the operations performed by the slice controller, and the slice utility gain, and optimizes (here maximizes) the overall utility gain of the network system (for example, by maximizing the sum of the slice utility gains of multiple network slices). In particular, in the present disclosure, the machine learning model is based on network parameters, decisions regarding the admission of slice requests, and data learning regarding the resulting slice utility gain, but at least by data learning regarding slice demand load, and can be understood as a learning algorithm that can optimize the overall utility gain of the network system.

[0030] Supplying a parameter (for example, slice demand load) to the machine learning model can be understood as the value of such a parameter being provided as an input to the machine learning model such that the predicted value (or estimated value) output by machine learning takes into account the value obtained for such a parameter.

[0031] Determining a decision regarding the permission of a slice request can be understood as an operation performed by the slice controller to permit or reject a slice request, that is, an operation to allocate or not allocate network resources to the corresponding user equipment. The permission of a slice request can be understood as providing the corresponding user equipment with the network resources required by such a user equipment according to the corresponding demand parameters for the requested service. The rejection of a slice request can be understood, for example, as buffering an incoming slice request or maintaining the buffering state of an incoming slice request.

[0032] According to the embodiments of the present disclosure, the following features can be optionally implemented.

[0033] In a possible embodiment, slice admission control includes performing a plurality of repetitions of at least steps a) to f), and two consecutive repetitions are spaced based on repetition parameters.

[0034] As a result, the proposed admission control method can be repeated for a plurality of incoming slice requests from a plurality of user equipment that request network resources from a plurality of network slices. Further, such a proposed admission control method can be repeated over time as the overall state of network resources in the network system (and thus the allocation of network resources in the network slice) and the incoming slice requests change. In particular, such repetitions update the machine learning model and optimize the slice admission control mechanism through the repetitions.

[0035] The iterative parameter can be understood as a parameter that leaves an interval from a certain iteration of steps a) to f) to the next iteration immediately following. For example, such an iterative parameter can be a time-related parameter such as periodicity (e.g., on the order of milliseconds, seconds, or minutes). In such a case, after the timing corresponding to such a time-related parameter has elapsed, another iteration of steps a) to f) is executed. In particular, the overall state of the updated network resources is obtained, and the parameters of the updated network slice (e.g., slice service variable, slice utility gain, slice demand load) are determined. Such an iterative parameter is received by the slice controller and is also related to the trigger signal that triggers the iteration. Such a trigger signal can be, for example, a signal received from a buffering unit that transmits a batch of slice requests to be processed by the slice controller. The iterative parameter can be the number of slice admission requests processed by the slice controller.

[0036] In a possible implementation form, the slice utility gain is determined based on the learning of at least the slice demand load supplied to the machine learning model in the previous iteration in the current iteration.

[0037] As a result, the proposed admission control method proposes a slice admission control mechanism based on the learning of specific parameters related to the impact of incoming slice requests on the resources given to the network slice. Therefore, with the proposed admission control method, slice requests for the network slice can be dynamically considered (and thus dynamically adjusted) based on the impact of the resources of those slice requests on the network slice. Through such learning, the user equipment can request network resources for a network slice with the smallest possible slice demand load, so that incoming slice requests can be formed and predictability can be introduced into the incoming slice requests.

[0038] In a possible embodiment, the slice demand load is obtained by the following formula.

Number

Number

[0039] As a result, the proposed admission control method introduces a slice demand load parameter to quantify the resource cost (or impact) of an incoming slice request for a given network slice. By knowing such slice demand loads for both the slice controller and the user equipment sending the slice request, an incoming slice request can be formed. By such request formation, on the one hand, the user equipment can determine the request method so as to have as many permitted slice requests as possible (for example, by distributing the demand parameters of those slice requests among various selected network slices), and on the other hand, the slice controller can predict and further anticipate the incoming slice request and the requested network slice.

[0040] The load parameter can be understood as a parameter related to the total amount of network resources in a plurality of network slices when at least one slice request from at least one user device is permitted. In other words, the load parameter can represent the maximum network resources available when the demands of all network slices are permitted. The load parameter can also be understood as the overall resources consumed by the network slice when all incoming demands of the network slice are accepted.

[0041] In a possible implementation form, the admission control method further includes, before obtaining the slice demand load, obtaining the value of the load parameter, and further includes. The value of the load parameter is updated as the following formula between two consecutive iterations.

Equation

Equation

[0042] As a result, by updating the value of the load parameter, the network system can improve resource efficiency at the application layer APP level.

[0043] In a possible implementation form, the admission control method supplies the load parameter to a machine learning model, and further includes. The slice utility gain is obtained based on the learning of at least the load parameter supplied to the machine learning model in the previous iterations in this iteration.

[0044] As a result, the proposed admission control method enables network resource load prediction and enables prediction of resource consumption of multiple network slices. Therefore, by obtaining the updated value of the load parameter throughout the entire continuous iteration and dynamically learning the value of the load parameter, the traffic in the deployment area can be predicted, for example, by securing more network resources for slice requests and increasing the resource efficiency of the system. The value of the load parameter is used to predict the resource consumption introduced by the slice request so that the slice controller can predict the next coming slice request based on such learning.

[0045] In a possible implementation form, the admission control method transmits the load parameter to at least the user equipment, and further includes.

[0046] As a result, by notifying the user equipment that sends a slice request of the load parameter value in the ongoing state of the network system by the proposed admission control method, it becomes possible to further optimize the request formation. Therefore, the user equipment can select a transmission method (and thus a request method) that optimizes load reduction across the network system. In fact, for example, based on the received slice demand load and the received load parameter, the user equipment can estimate the slice utility gain associated with each network slice. Since the user equipment can select a less loaded network slice in a predictable manner for its own transmission (and thus its own slice request), the network slice admission control mechanism can then be further optimized.

[0047] In a possible embodiment, after a predetermined number of iterations, the slice admission control is performed based on a sub - part of the plurality of network slices, and the sub - part is composed of network slices where the corresponding slice demand load is less than a predetermined load threshold.

[0048] As a result, the proposed admission control method can reduce the computational complexity of the slice admission control mechanism and shorten the convergence time of the machine learning model that enables the operation of the slice admission control mechanism. For example, in each iteration, by considering only sub-parts of multiple network slices, the slice controller can discard some network slices with higher slice demand load (i.e., higher than a predetermined load threshold) when performing slice admission control. Therefore, the admission control method can perform state space reduction because the number of network slices considered for permitting slice requests is reduced with respect to the total number of network slices deployed in the network system. In fact, in the existing slice admission mechanism, the Q-value (i.e., the value for selecting an operation to optimize the overall network resources of the system in each state of the network system) is learned based on a large state space and / or operation space of a potentially large number of iterations before all Q-values are learned. Thus, the main problems are the computational complexity and convergence time of the learning (e.g., reinforcement learning) model.

[0049] In the proposed admission control method, the introduction and dynamic update of slice demand load and their transmission to user equipment enable the prediction of the network slices required in the next slice request sent by such user equipment.

[0050] The sub-parts of multiple network slices can be understood as some of the multiple network slices, and the number of network slices in such sub-parts is less than the number of multiple network slices. Such sub-parts of network slices may change between two consecutive iterations.

[0051] The predetermined load threshold can be understood as a preconfigured value corresponding to the slice demand load (e.g., represented by resource blocks per throughput unit). Such a load threshold can be, for example, the average or median of the slice demand loads of all network slices. Such a load threshold can also be configured by the slice controller according to, for example, the number of multiple network slices of the network system or the size of the resource pool.

[0052] After a predetermined number of iterations can be understood as reaching an iteration where the change in slice demand load between two consecutive iterations is less than a specific threshold predefined, for example, as 1% of the initial slice demand load. After a predetermined number of iterations can also be understood as the number of iterations when the average change in slice demand load over a specific period is below a specific threshold, for example, 10% of the initial average value of the slice demand load at the start of such a period. After a predetermined number of iterations can also be understood as a fixed number of iterations after which a stable slice demand load is considered to be obtained, for example, after 3, 5, or 10 iterations (stable means that the change in slice demand load is understood to be less than a predefined threshold). Such thresholds and the number of iterations depend on the assumptions of the network slice deployment.

[0053] In a possible implementation, the decision to permit a slice admission request is based on optimizing the service utility functions of multiple network slices, and the above service utility function is represented by the following formula.

Number

Number

[0054] As a result, the proposed admission control method models the slice admission control mechanism as an optimization problem considering the slice demand load and slice service variable defined above. Therefore, the decision regarding the admission of an incoming slice request (i.e., the decision to permit or reject) can be determined by optimizing the service utility function of the network system considering the resource constraints of such a system.

[0055] The service utility function can be understood as a function that reflects the overall utility of network resources in the entire network system. Such a service utility function can be understood, for example, as the sum of all slice utility functions of multiple network slices.

[0056] In a possible implementation form, before step d), the admission control method transmits a gain check signal to the application layer, and further includes Such a gain check signal includes data related to a change in at least one slice utility gain considering updated values of resource budget parameters.

[0057] As a result, the proposed admission control method makes it possible to further optimize the slice admission control mechanism by considering the relaxation of resource constraints in the network system. In fact, the admission control method proposes to utilize the potential flexibility of the application layer in order to relax (i.e., change) the budget parameters configured by the application layer for a plurality of network slices.

[0058] After that, by relaxing (i.e., changing) such resource budget parameters, further service utility gains can be considered, and accordingly, updated values of the resource budget parameters can be considered.

[0059] In a possible implementation form, the change in the at least one slice utility gain considering the updated value of the resource budget parameter can be expressed, for example, by the following formula.

Number

Number

[0060] In a possible embodiment, the admission control method based on the response signal received from the application layer, obtains an updated value of the resource budget parameter during two consecutive iterations as follows, that is,

Equation

[0061] According to another aspect of the present disclosure, a slice controller configured to execute a slice admission control mechanism in a communication network system, where the communication network system deploys a plurality of network slices, and such a slice controller includes at least a processing unit, and a non - transient computer - readable medium storing instructions that configure the slice controller to execute the admission control method when executed by the processing unit, and is provided. A slice controller is also proposed.

[0062] According to another aspect of the present disclosure, an admission request method executed by a user equipment for sending a slice admission request for at least one selected network slice in a communication network system that deploys a plurality of network slices, the slice admission request includes at least data related to slice request parameters, and the admission request method includes: Receiving slice demand loads respectively associated with a plurality of network slices, each slice demand load being related to the network resources consumed in the corresponding network slice when the slice admission request is permitted; Adapting the slice request parameters of the slice request so as to request network resources for at least one selected network slice among the plurality of network slices based on at least the received slice demand loads, the selected network slice having a corresponding slice demand load less than a predetermined load threshold; Sending a slice admission request based on the adapted slice request parameters; An admission request method including the above is also proposed.

[0063] As a result, the proposed admission request method enables assisting the user equipment when sending a slice admission request to the network system. Therefore, by receiving at least the slice demand loads of the network slices, the user equipment can estimate the change in network utility, locally evaluate the slice resource cost of each network slice, and select the network slice with the lowest resource cost (i.e., the lowest slice demand load) for their transmission (thus, for obtaining the demand parameters of their slice admission requests).

[0064] According to another aspect of the present disclosure, a user equipment configured to send a slice admission request for at least one selected network slice in a communication network system that deploys a plurality of network slices, at least, a user processing unit, and a non-transitory computer-readable medium for a user device, including stored instructions that configure the user device to execute an admission request method when executed by the user processing unit, is also proposed.

[0065] According to another aspect of the present disclosure, a computer program product including program instruction code stored in a computer-readable medium that executes an admission control method is also proposed.

[0066] According to another aspect of the present disclosure, a computer program product including program instruction code stored in a computer-readable medium that executes an admission request method is also proposed.

[0067] According to another aspect of the present disclosure, a non-transitory storage medium readable by a processor, which stores a program for implementing an admission control method when a computer program is executed by the processor, is also proposed.

[0068] According to another aspect of the present disclosure, a non-transitory storage medium readable by a user processor, which stores a program for implementing an admission request method when a computer program is executed by the user processor, is also proposed.

[0069] Other features, details, and advantages are shown in the following detailed description and the figures.

Brief Description of the Drawings

[0070]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Mode for Carrying Out the Invention

[0071] Refer to FIG. 1. FIG. 1 schematically shows a wireless communication network system SYS. Such a system SYS is defined for a fifth generation (5G) wireless network. The architecture of the system SYS is mainly composed of different layers including an infrastructure layer, a control layer, and an application layer.

[0072] The infrastructure layer includes a deployment area DA composed of a plurality of network devices that transfer data and resources of the network system SYS. Such network devices include 5G access network base stations BS that form a radio access network (RAN). Such a RAN can transfer data and signals between user equipment (UE) UE1 to UE12 and an application layer user plane APPU using both a shared access network SAN and a transport layer TN that are both composed of a 5G core network layer CNW.

[0073] The 5G core network CNW is part of the control layer. The control layer enables data transfer, resource allocation, network topology control, information collection, and the configuration and setup of the transfer network TN in the network system SYS. The core network CNW manages the shared access network SAN in the deployment area DA. The management of such a shared access network SAN is carried out by the access mobility management function AMF of the core network CNW. Such a function AMF guarantees the access of user devices UE1~UE12 to the 5G base station BS and executes resource allocation and mobility management in the deployment area DA. The session parameters across the transfer network TN are controlled by the session management function SMF and coordinated with the resource allocation and mobility management provided by the access mobility management function AMF.

[0074] The application layer includes multiple application services that use the resources of the network system SYS. These resources are determined by the core network CNW and are exposed to the application control function APPC by the network exposer function NEF. The exposure of resources by the network exposer function NEF to the application control function APPC may refer to the transmission of resources to the application control function APPC for cloud type management purposes. For example, the network exposer function NEF can map the resources of the shared access network SAN and the transfer network TN to cloud resources such as networking resources, storage resources, and computing resources. Cloud resources are essentially virtual machine (VM) resources. The application layer is essentially an application server APPU that collects and / or transmits application-related data from the deployment area DA and the application layer control function APPC, and is responsible for determining the service requirements of the application services operating on the network system SYS (therefore, the service requirements of the UE). Such service requirements define the parameters of the application layer communication between the application server APPU and the user equipment UE1 to UE12 in the deployment area DA. Such parameters can include, for example, application layer throughput, application layer delay, application layer service reliability, or any other parameter that can describe the application layer communication in the deployment area DA. The application layer server APPU and the application layer control function APPC are arranged in the edge network EC where the application layer control function APPC can host the application control functions of multiple applications.

[0075] The user plane of the application layer (i.e., the application server) APPU is connected to the infrastructure layer, while the control plane (i.e., the application layer control function) APPC uses the resource information disclosed by the network exposer function NEF and determined by the core network CNW.

[0076] Regarding the present disclosure, the network system SYS is supported by a software-defined network (SDN) architecture and a network function virtualization (NFV) architecture. In particular, by combining SDN and NFV, the 5G network system SYS can abstract multiple independent virtual networks based on a single, aforementioned physical infrastructure of the network system SYS. Such an architecture enables the deployment and management of multiple network slices. The division of such virtual networks is called network slicing. As a result, in the present disclosure, the deployment of the 5G wireless communication network system SYS is considered to involve the deployment and management of multiple network slices.

[0077] Therefore, each network slice refers to an independent virtual end-to-end network capable of resource allocation and process optimization to provide a service type suitable for a specific purpose. Thus, through network slicing, the network system SYS can adapt network functions to efficiently support a wide range of service types with various quality of service (QoS) requirements, as described above, using a single physical infrastructure and a common physical resource pool. Each network slice uses a part of the radio access network and deploys network functions suitable for the purpose to ensure the transfer and support of slice-related data and control packets with slice-specific QoS requirements.

[0078] Some network slices can be designed to support enhanced mobile broadband (eMBB) such as high speed and high bandwidth applications. Other network slices can be designed to support massive machine type communication (mMTC) such as IoT use cases. Other different network slices can support ultra-reliable low-latency communication (URLLC) such as critical infrastructure communication. Referring to Figure 1, multiple user equipments UE1 to UE12 are deployed on a 5G network system SYS using different network slices. For example, user equipments UE1, UE5, UE10, UE12 can be supported by a common first (network) slice targeted at supporting large-scale broadband applications. Such user equipments UE1, UE5, UE10, UE12 have a common service type, such as high-definition video streaming, and thus similar QoS requirements (QoS requirements regarding throughput). User equipments UE2, UE6, UE7, UE9, UE11 can be supported by a common second (network) slice different from the first slice, and such a second slice is targeted at supporting large-scale communication. Such user equipments UE2, UE6, UE7, UE9, UE11 have a common service type, such as remote monitoring or remote communication, and thus common QoS requirements (QoS requirements regarding energy efficiency and UE density). User equipments UE3, UE4 and UE8 can be supported by a common third (network) slice different from the first slice and the second slice, and such a third slice is targeted at supporting ultra-reliable low-latency communication. Such user equipments UE3, UE4, UE8 have a common service type, such as driving assistance, and thus common QoS requirements (QoS requirements having high reliability, fast response, and the required security level).Therefore, the capabilities provided by each network slice depend on the service class of such a network slice, i.e., the type of service that the network slice intends to provide in terms of technical performance (delay, throughput, service continuity, speed, security, etc.).

[0079] Therefore, multiple (network) slices operating simultaneously on the same network infrastructure mean that the specifications of resources and network functions are heterogeneous for different slices. Such slice specifications are managed by the application layer APPC.

[0080] The 5G wireless communication network system SYS receives user requests (or demands) that support various service types required by various user devices of the network system SYS. Such user requests are equally processed by the system SYS as slice requests aimed at being allocated together with network resources corresponding to network slices. For example, a user request may refer to a throughput requirement in a given network slice. To avoid resource access collisions within the network system SYS, incoming user requests are scheduled and processed by the network system SYS. Therefore, the network system SYS deploys a slice admission control mechanism (or policy) to determine the approval or rejection of each incoming slice request. Such a slice admission control policy is executed by the slice controller SC of the system SYS. The slice admission control policy aims to optimize the long-term overall network performance by satisfying as many service and user device requirements as possible while considering the available resources of the system SYS, the capabilities of the system SYS, and the specifications to be satisfied by user devices already provided by the slice.

[0081] Next, refer to FIG. 2. FIG. 2 illustrates a slice admission control mechanism executed in a 5G communication network system SYS. Such a network system SYS is shown, for example, in FIG. 1. The slice admission control mechanism can be executed by a slice controller SC of such a system SYS.

[0082] The purpose of the slice admission control mechanism is to determine a long-term optimal policy in slice request allocation. In other words, the slice admission control mechanism is for the slice controller SC to determine whether to permit or reject each incoming user slice request so as to maximize the network utility function. Such a network utility function can generally be defined by the following formula.

Equation

[0083] Therefore, the slice admission control mechanism can be modeled as a Markov Decision Process (MDP) in the environment of the 5G communication network system SYS as schematically shown in FIG. 2.

[0084] In step P0, the network communication system SYS receives an incoming user request. Such an incoming user request can be received by the request buffer BF of the system SYS. In one embodiment, the request buffer BF can be part of the slice controller SC. The request buffer BF stores the incoming user request. Such an incoming user request can be stored according to a queueing model. The user request is buffered by the request buffer BF such that such requests are processed by the slice controller SC in the order of arrival of the incoming user requests. The request buffer BF can also preprocess the incoming user request based on the service type and quality of service (QoS) requirements of such requests. Therefore, the request buffer BF can convert the incoming user request into a slice request. In other words, the request buffer BF can sort the incoming user request based on the network slice and the corresponding network resources allocated to such requests. The request buffer BF can buffer such slice requests in one or several queues, for example, according to the requested service. In the remainder of this specification, "user request" and "slice request" are used interchangeably.

[0085] In step P1, the slice controller SC receives from the request buffer BF information related to the buffered slice requests. Such information can be timestamped. Such information can include a batch of user requests and can be received periodically according to a preconfigured period. Such information can also include different request queues and can be received by the slice controller SC according to different preconfigured periodicities. Also, such information can include a buffer status indicating the number of buffered slice requests or the level of the queue length. The number of such buffered slice requests can be indicated for each queue. Such information can include data related to each buffered slice request. Such data related to each buffered slice request can include, for example, the following. Data related to the arrival time of each incoming slice request, Data related to the buffering time of each buffered slice request (i.e., the time when each slice request is buffered by the request buffer BF), Data related to the delay requirement (i.e., the level of delay tolerance) of each buffered slice request.

[0086] Based on such information received from the buffer BF associated with the time under consideration, the slice controller SC can determine the buffering cost L of the system SYS associated with such time. del In other words, based on the timestamped information received from the request buffer BF, the slice controller SC can quantify the service delay, i.e., the queueing delay of the system SYS due to an incoming slice request being buffered in the request buffer BF and waiting for processing by the slice controller SC. Such a buffering cost L del can be directly determined by the request buffer BF and can also be sent to the slice controller SC in step P1.

[0087] In step P2, the slice controller SC receives network information related to the current state of the communication network system SYS. Such network information can be provided by the network subsystem SSYS. Such a subsystem SSYS manages the wireless access network of the system SYS when transferring and sharing resources to the deployment area DA of the system SYS. Such a subsystem SSYS shares the resources of the system SYS for various services currently supported by the network slices of the network system SYS. Such a subsystem SSYS can include, for example, the core network CNW, edge computing EC, shared access network SAN, and transfer network TN of the system SYS.

[0088] Such network information can be timestamped (i.e., associated with the current time) and can be related to, for example, the following. The total amount of available resources in the system SYS at the current time and / or at a time preceding the current time (e.g., in units of resource blocks), The available amount of slice resources of the system SYS in different network slices at the current time and / or at a time preceding the current time, The service types and requirements currently satisfied by the system SYS at the current time and / or at a time preceding the current time, and / or The current values of QoS parameters (delay, reliability, throughput, etc.) in each network slice at the current time and / or at a time preceding the current time.

[0089] Such step P2 can be executed simultaneously with step P1. Therefore, the received information about the resource state of the network system SYS and the received information about the request buffer BF have similar or close timestamps (i.e., are associated with those timestamps).

[0090] Based on such information received from the subsystem SSYS and the request buffer BF, the slice controller SC can obtain observation information regarding at least the current state of the network system SYS.

[0091] In step P3, the slice controller SC processes the incoming slice requests currently buffered in the request buffer BF based on the information related to the buffered slice requests received in step P1 and the information related to at least the current state of the network system SYS received in step P2, i.e., the current state of the network system SYS. The slice controller SC can process the incoming slice requests according to the first-in-first-out (FIFO) processing of the queue in the request buffer BF. The slice controller SC can also process the incoming slice requests according to the request parameters or priority levels of the incoming slice requests. The processing of the incoming slice requests is to determine the decision (i.e., operation) a of permission or rejection of the received slice requests. The determination of the permission or rejection operation a of the received slice requests can be performed by the slice admission control module SAC MD included in the slice controller SC.

[0092] The decision a of permission or rejection of the processed incoming slice requests is associated with the current state of the network system SYS as observed by the slice controller SC based on steps P1 and P2. The determination of such a decision a can depend on a reinforcement learning-based algorithm or a deep reinforcement learning-based algorithm executed by the slice admission control module SAC MD. Such learning algorithms that result in the decision a made according to the current state will be described in more detail herein.

[0093] In step P4, the decision a of the slice controller SC is sent to the request buffer BF. If the decision a of the slice controller SC permits the processed slice request, and thus responds to the processed slice request, the request buffer BF can be updated in that the permitted slice request is removed from the queue of the request buffer BF and the buffer status of the request buffer BF is updated accordingly. On the other hand, if the decision a of the slice controller SC rejects the slice request, the request buffer BF can hold the processed slice request in the queue of the request buffer BF. The buffer status of the request buffer BF can also be updated (for example, by increasing the buffering time of the rejected slice request). The buffering cost L del may also increase.

[0094] In step P5, at least the decision a of the slice controller SC that permits or rejects the processed slice request is sent to the subsystem SSYS to update the current state st of the network system SYS. Such a decision a can be sent to the subsystem SSYS by the request buffer BF or the slice controller SC. The request buffer BF can also send the data related to the processed slice request in step P5. The decision a made by the slice controller SC may affect the overall state and / or utility of the network system SYS. In fact, if the decision a of the slice controller SC permits the processed slice request, the subsystem SSYS can allocate the resources of at least one network slice allocated to the permitted slice request. In such a case, the amount of resources available in the network slice changes and the overall state of the network system SYS changes. The utility gain G of the network system SYS adm increases due to the decision a that permits the processed slice request.

[0095] In step P6, a reward r triggered by the subsystem SSYS is brought about by the decision a determined by the slice controller SC in step P3. Such a reward r is sent to the slice controller SC (or the slice admission control module SAC MD). The reward r is due to the action a taken by the slice controller SC according to the state st of the network system SYS. Such a reward r can be, for example, a positive integer value when the decision a permits the slice request being processed, and can be zero or a negative integer value when the decision a rejects the slice request being processed. In a particular embodiment, the reward r can be different according to the data related to the processed slice request. For example, when a high-priority service type is required for the processed slice request, or there are strict service quality requirements, a higher reward r can be obtained by such permission compared to the permission of a slice request with high delay tolerance.

[0096] Steps P0 to P6 represent one iteration in the slice admission control mechanism executed by the slice controller SC, particularly the slice admission control module SAC MD. Thus, such an iteration is to determine the action a of permitting or rejecting an incoming slice request according to the current state st of the network system SYS at a given time t. Such an action a generates a reward r.

[0097] More generally, the Markov decision process of the slice admission control mechanism executed by the slice controller SC of the 5G communication network system SYS can be modeled by the following. The state space S of the network system SYS. This is understood as the set of all possible states st of all network slices of the network system SYS at a given time. For example, the possible values of the amount of resources available in the network slice, throughput, delay, reliability, and / or any other quality of service parameter in all network slices measured at a given time can form the set of all possible states st of the network system SYS. In the present disclosure, the state space S is considered to be finite or infinite. The operation (or decision) space A of the slice controller SC. This is understood as the set of decisions that may be made by the slice controller SC. In the present disclosure, the operation space A can have a dimension (or size) equal to, for example, 2. The operation space can include boolean values or integer values corresponding to the permission or rejection of incoming slice requests, respectively. The reward value r. This is understood as the gain value or loss value obtained by the slice controller SC after the operation a is performed according to the given state st.

[0098] The slice admission control mechanism consists of repeatedly executing steps P0 to P6 several times to maximize the network utility function Q. In each iteration, considering the state st and the subsequent operation a performed, the network utility function Q is updated. That is, the utility gain G adm may change (for example, when a positive reward r is obtained, the utility gain G adm increases), and / or the buffering cost L del may change (for example, when a processed slice request is rejected, the buffering cost L del increases).

[0099] The slice admission control mechanism consists of updating the network utility function Q based on the learning process of the state st and the subsequent operations and rewards respectively associated with such state st.

[0100] In the first embodiment, the slice admission control mechanism uses a reinforcement learning-based algorithm, particularly a Q-learning-based algorithm. A detailed description of the reinforcement learning algorithm is described, for example, in the prior art literature, Nguyen, Dinh et al., "Application of Deep Reinforcement Learning in Communications and Networking: A Survey", (Part II, section B).

[0101] The slice admission control mechanism using reinforcement learning (more specifically, Q-learning) consists of exploring a table, also called a Q-table, that encompasses all pairs of states and actions (st,a) that occur in the system SYS. The pairs of states and actions (st,a) that occur in the system SYS are associated with the maximized utility function Q obtained by iterating learning over time. More precisely, for a given time t+1 corresponding to, for example, the time when a new slice request is received (in other words, when the second iteration of steps P0 - P6 is performed), the network utility function Q of a specific pair of state and action (st,a), also called the Q-value, is updated as follows.

Equation

[0102] Time t and time t + 1 can respectively correspond to the time slots in two consecutive iterations of steps P0 to P6 (i.e., the time of two consecutive iterations of the Q - learning method).

[0103] Learning rate α t And the initial values of the discount rate γ can be predefined or pre - configured in the slice admission control module SAC MD.

[0104] As a result, after exploring all pairs (st, a) of states and actions in the Q - table, the slice admission control mechanism can obtain the mapping of all states of the network system SYS in the state space when all Q - values converge or after a certain number of iterations of steps P0 to P6 have elapsed. Therefore, the Q - values obtained by reinforcement learning provide an optimal policy for slice request allocation. Thus, when the slice controller SC receives a new slice request at time T, the slice controller SC determines the mapped action a to reject or permit such an incoming slice request based on the observed state st of the network system SYS at time T, and can maintain or reach the optimal Q - value Q (st, a). * based on, and the mapped action a to reject or permit such an incoming slice request * to determine, and the optimal Q - value Q T (st * , a * ) can be maintained or reached.

[0105] In the second embodiment, the slice admission control mechanism uses a deep reinforcement learning-based algorithm, particularly a deep Q learning-based algorithm (DQL (deep Q learning-based) algorithm). A detailed description of the deep reinforcement learning algorithm is described, for example, in the prior art document, Nguyen, Dinh et al., "Application of Deep Reinforcement Learning in Communications and Networking: A Survey", (Part II, section B).

[0106] The slice admission control mechanism using a deep reinforcement learning-based algorithm consists of implementing a deep Q network (DQN) instead of a Q table to maximize the network utility function. Therefore, such a mechanism consists of executing a combination of reinforcement learning and deep learning.

[0107] Reinforcement learning means exploring and learning a Q table to maximize the value of the network utility function Q for each pair of state and action (st,a). However, the computational complexity of such exploration depends on the size and / or complexity of the state space S and action space A of the system SYS. For example, in the case of a large action space A and state space S (for example, a state space and / or action space with more than 10 dimensions), the slice admission control mechanism using reinforcement learning may not be able to obtain the optimal policy for slice admission control within a reasonable time (for example, less than 1000 iterations of the Q learning method). In other words, the convergence time of the reinforcement learning algorithm of the slice admission control mechanism may be slow.

[0108] Furthermore, in Q learning, due to the fact that the samples of states and actions used to estimate the optimal action are the same as the samples of states and actions used to obtain each action value in each iteration, there is often a bias towards the action determined to be optimal considering the pair of state st and action a.

[0109] Therefore, deep reinforcement learning can be executed using a replay memory to replay the states and actions and the subsequent rewards obtained by the network system SYS. Also, deep reinforcement learning can be executed by training two different neural networks (and thus two different Q-value functions) to simultaneously select and evaluate the actions taken in each iteration. Such an approach is called double deep Q-learning.

[0110] Next, refer to FIG. 3. FIG. 3 is a flowchart showing the steps of executing a slice admission control mechanism in an embodiment proposed in the present disclosure.

[0111] The slice admission control mechanism executed by the slice controller SC depends on an admission decision for a UE slice request in order to maximize the overall utility function that provides a constrained environment for the 5G communication network system SYS. In fact, the network system SYS is constrained by a constraint function that depends on the available amount of resources in the network slice, service parameters, specifications set by the edge computing level EC, and service requirement parameters.

[0112] Existing slice admission control mechanisms use reinforcement learning algorithms and deep reinforcement learning-based algorithms to maximize the network utility function Q. However, the computational complexity and convergence time of such learning algorithms are due to the unpredictable and random nature of the incoming user requests (e.g., user requests to access the network slice with a specific throughput) assumed in existing slice admission control mechanisms, the fixed constraint functions to be complied with, and remain a major issue in the slice admission control mechanism.

[0113] In this disclosure, it is assumed that a 5G communication network system includes N network slices. Such network slices can provide various service types, and such service types may potentially have various service quality requirements regarding, for example, various priorities, latencies, throughputs, and reliabilities. Corresponding resources are allocated to the various network slices. User equipments UE1 to UE12 send user requests to the network system SYS. Such user requests may refer to the demand for traffic (or throughput) in one or more network slices according to the data packets transmitted regarding the services required by the user equipments UE1 to UE12. For example, the user requests sent to the slice controller SC (or request buffer BF) of the network system SYS may be periodic requests that require a minimum level of throughput and / or a maximum level of tolerance for latency and / or a minimum level of reliability for data packets transmitted across the network system SYS, or specific time-critical requests with specified service quality parameters. It can be.

[0114] Considering such incoming user requests, this disclosure defines a slice service variable t s parameter associated with each of the network slices of the network system SYS. In other words, for any index value s from 1 to N, the slice service variable t s is associated with the network slice s. The slice service variable t s can refer to, for example, the overall throughput required by the user across the network slice s. In other words, t s can refer to the sum of the throughputs of the user equipments accessing the network slice s. In another embodiment, the slice service variable t scan be the total throughput respectively required by user equipment requesting access to network slice s. In yet another embodiment, service slice variable t s is the average throughput of user terminals accessing network slice s, or the average requested throughput of user terminals requesting access to network slice s. More generally, slice service variable t s can be understood as a parameter reflecting the updated state and quality of service parameters (throughput, reliability, security, latency, etc.) that can be provided by such corresponding network slice s when the corresponding network slice s are allocated to incoming user requests.

[0115] When the incoming slice request requests traffic in one or more network slices, slice service variable t s can be understood as the total throughput provided by at least one network slice in the deployment. In other words, slice service variable t s is a parameter reflecting the impact of the incoming slice request on the quality of service parameters provided by each network slice.

[0116] As a result, the service utility function of network system SYS maximized by slice controller SC when executing the slice admission control mechanism can be defined as the following multivariable function.

Equation

[0117] Such a service utility function to be maximized represents the overall utility function of the network system SYS, that is, the utility functions of all network slices of the system SYS. In other words, such a service utility function represents the average service gain obtained from various user requests received by the network system SYS.

[0118] The utility function U of each network slice s s is represented by the following formula.

Equation

[0119] When considering the network slice s, such a utility function U s can be represented as an example by the following formula.

Equation

[0120] In the present disclosure, the introduction of the slice demand load p s parameter is also proposed. The slice demand load p scan refer to the estimated impact of user demand on network slice s, and such user demand can be, for example, the slice service variable t regarding the resource consumption of network slice s s represented by. In other words, the slice demand load p s is the estimated load across network slice s that would occur if the demands of all users generating the combined throughput (or slice service variable) of t s across network slice s were accepted. The larger the slice demand load p s for network slice s, the greater the resource consumption of the user requirements represented by the slice service variable t s . Similarly, the smaller the slice demand load p s for network slice s, the smaller the resource consumption of the user requirements represented by the slice service variable t s .

[0121] In other words, for a given network slice s, the slice demand load p s can be understood as the resource cost for permitting the incoming slice requests to be considered in network slice s. Therefore, a network slice s with a high value of slice demand load p s reflects a network slice that consumes a high amount of network resources compared to another network slice s' with a lower value of slice demand load p s’ . If network slice s cannot respond to the incoming slice requests, its corresponding slice demand load p sis zero. A network slice that requires a large amount of resources (for example, because such a network slice provides a service that requires a large amount of resources) may cause a resource shortage for other services provided by other network slices when resources for network slices that require a large amount of resources are allocated to incoming user requests. The slice demand load p s of any network slice can be defined as the amount of physical resource blocks per slice service variable t s . For example, when the service provided by a network slice provides throughput to data packets, the slice service variable t s to be considered for this network slice is throughput, and the slice demand load p s is the amount of physical resource block cost to the system SYS per throughput in such a network slice. The slice demand load p s of each network slice depends on the buffer status, access network status, and / or transfer network status regarding the resources of each network slice.

[0122] The network system SYS, more specifically the application layer (here, the control plane of the application layer APPC), sets the total amount of resources allocated to various network slices. The total amount of resources may be referred to as the resource budget I parameter. The application control function APPC can set, for example, a parameter derived from the amount of throughput of a network slice. The application control function APPC can then send that parameter to the pivotal cloud fundry PCF of the core network CNW. The pivotal cloud fundry PCF can further send that parameter to the access mobility management function AMF and the session management function SMF, resulting in session resources and access resources being obtained.

[0123] Therefore, the resource budget I can be understood as, for example, cloud or processing resources. Therefore, the resource budget I parameter refers to the amount of physical resource blocks reserved for network slicing at the edge computing EC level.

[0124] When the network system SYS receives an incoming slice request, the slice admission control mechanism is executed by the slice controller SC considering the following constraint functions. [Number] Here, p s is the slice demand load of the network slice s. t s is the slice service variable of the network slice s. I is the resource budget allocated to all network slices of the system SYS. N is the number of network slices in the system SYS.

[0125] As described above, the slice demand load p of the network slice s s can be understood as the resource load introduced by the incoming demand. The slice service variable t s can be understood as the throughput of the network slice s shared by the user equipment requesting resources for the network slice s. The resource budget I can be understood as the overall throughput available across the network system SYS.

[0126] For example, at the initial time of the slice admission control mechanism (e.g., the start time of the disclosed slice admission control mechanism), the initial value of the slice service variable t s is t1 = t2 =... = t Nis set to be \(I / N\), and the slice demand load \(p\) s has an initial value of \(p_1 = p_2=\cdots=p\) N is set to \(\alpha\lt1\). At each iteration of the slice admission control mechanism and / or, for example, after each admission of a slice request, such a slice demand load \(p\) s and the slice service variable \(t\) s are updated. Such an update depends on the derivative of the utility function \(U\) s associated with each network slice \(s\). Since such a utility function \(U\) s is predictable or estimated using a model learning algorithm such as a (deep) reinforcement learning algorithm, the slice demand can be predicted using such a learning algorithm. As a result, the overall slice admission control mechanism can use only one model learning algorithm to predict the future state and parameters of the network system SYS.

[0127] As a result, in the present disclosure, the slice admission control mechanism can be modeled as the following constrained optimization problem.

Equation

Equation

[0128] In fact, such an optimization problem will be considered by the slice controller SC when processing each received user request in order to quantify the overall performance gain (or loss) of the network system SYS when the incoming user requests are permitted by the slice controller SC.

[0129] Such an optimization problem will be solved in each iteration of the slice admission control mechanism. For this purpose, the Lagrange method using the Kuhn-Tucker conditions can be used. Therefore, the Lagrange function L can be defined as the following formula.

Equation

[0130] The maximization of the utility function U s under the constraint function is obtained by a specific value of the slice service variable t s However, such a specific value is conditional on satisfying the following formula for any index value s within the range of 1 to N. [Number]

[0131] In the present disclosure for solving the aforementioned optimization problem, the slice utility function U of network slice s s is approximated as a function that depends on the slice load parameter t associated with such a network slice s. This slice utility function satisfies ∂U s / ∂t s such that it reaches the maximum value of the derivative of the slice utility function U s (i.e., the utility gain of network slice s). In other words, this slice utility function is approximated to satisfy the following equation for any network slice s. s [Number]

[0132] Therefore, the optimization problem can be expressed as the following equation. [Number]

[0133] Value [Number] corresponds to the gradient (i.e., change) of the utility function of network slice s on the condition that the incoming user request is permitted by the slice controller SC and allocated to slice s. In other words, such a gradient value represents either a gain ( [Number] when it is positive) or a loss ( [Number] when it is negative) obtained by the network system SYS.

[0134] ​ When the optimization problem is solved considering all network slices (i.e., considering the gradient values of the utility functions obtained for all slice service variables t1, t2, t3,..., t N ), the ratio between such gradient values and the corresponding slice demand load is constant and corresponds to the load parameter λ.

[0135] Therefore, the slice demand load p of any network slice m s can be expressed by the following equation.

Equation

[0136] Furthermore, the load parameter λ can also be expressed by the following equation.

Equation

[0137] In other words, the load parameter λ quantifies the impact of the change in the resource budget I on the utility gain or loss of the network system SYS (e.g., by relaxing or suppressing the resource budget I set by the application layer APPC of the network slice). Therefore, when solving the optimization problem in a given iteration of the slice admission control mechanism, the slice controller SC can obtain the overall gain or loss of the utility of the network system SYS considering the relaxation of the constraint function

Equation

[0138] The load parameter λ is related to the slice demand load p sIt can be used to obtain. Such a load parameter λ can be updated based on the difference between the amount of resources associated with the permitted requests and the current resource budget I of the network system SYS.

[0139] Next, the step of executing the slice admission control mechanism according to an embodiment of the present disclosure shown in FIG. 3 will be described in detail. In such an embodiment of the present disclosure, the slice controller SC optimizes the slice admission control mechanism by forming an incoming user request. The formation of such user requests aims to address the drawbacks of existing slice admission control mechanisms, particularly the unpredictable and random nature of the user requests received by the slice controller SC.

[0140] In step S1, the slice controller SC receives at least one user request from at least one user equipment deployed in the network system SYS. Such a step S1 can correspond, for example, to steps P0 and P1 in FIG. 2. The slice request includes data related to slice request parameters that indicate service parameters and quality of service parameters required by the user equipment. For example, the user request received in step S1 can include the throughput required by the user.

[0141] In step S2, the slice controller SC determines the values of the slice service variables t1, t2, t3,..., t N for all network slices. The values of such slice service variables t1, t2, t3,..., t N can be obtained based on the slice request parameters related to the user request received in step S1 and a measure related to the state and available amount of the resources of the network slice.

[0142] In step S3, the slice controller SC determines the utility function U for all network slices ss to obtain (where s is, for example, from 1 to N, and N is the number of network slices). The slice controller SC determines the utility gain (or loss) function

Number

Number

[0143] In block step B1, the slice controller SC proceeds to determine the slice demand load method to form an incoming user request. To do so, the slice controller SC obtains the value of the load parameter λ in step S4. When the slice admission control mechanism is first executed (i.e., in the initial iteration) by the slice controller SC, the initial value λ k=0 of the load parameter λ can be set or preconfigured, for example, by the slice controller SC and / or the application control function APPC. Such an initial value λ k=0 can be set, for example, to λ k=0 = 1. The existing λ of the load parameter λ k>0 ​When the value is stored by the slice controller SC, for example, in the storage memory MEM of the slice controller SC (especially when the previous iteration k of the slice admission control mechanism has already been executed), the slice controller SC can use such a value of the load parameter λ. In step SS4, the value of the load parameter λ used can be supplied to the deep reinforcement learning model DQL.

[0144] In step S50, the slice controller SC determines the slice demand loads p1, p2, p3,..., p of various network slices N of the values, for each slice service variable t1, t2, t3,..., t N by dividing the change in the utility function with respect to the load parameter λ set in step S4. In other words, for any index s in the range from 1 to N, the slice demand load p of the network slice indexed by s s is determined by the slice controller SC as follows.

Equation

Equation

[0145] In step SS50, such values of the slice demand loads p1, p2, p3,..., p determined by the slice controller SC for various network slices N are such slice demand loads p1, p2, p3,..., p NIt can be supplied to the deep reinforcement learning model DQL for learning related thereto.

[0146] In step S70, the slice controller SC transmits the values of the slice demand loads p1, p2, p3, ..., p N to the user equipment UE1 to UE12 deployed in the deployment area of the network system SYS. Such transmission can be performed via the core network CNW and the shared access network SAN of the network system SYS. Thereby, data related to the slice demand loads p1, p2, p3, ..., p N can be signaled by the base station BS to the user equipment UE1 to UE12 as, for example, a broadcast signal, a multicast signal, and / or a unicast signal.

[0147] In step S70, for the slice demand loads p1, p2, p3, ..., p N transmission of the values enables the formation of incoming user requests in the long term for future iterations of the slice admission control mechanism. In fact, based on the values of the slice demand loads p1, p2, p3, ..., p N each user equipment UE1 to UE12 updates or adapts its future slice requests to the demand resources, etc. from the network slice having the lowest value among the slice demand loads p1, p2, p3, ..., p N For example, following the reception of the slice demand loads p1, p2, p3, ..., p N if the corresponding network slice s provides a throughput service and has a slice demand load p s exceeding a predetermined load threshold, the user equipment UE3 can reduce the throughput parameter of the service level specification in the next slice request of the network slice s. Such a predetermined load threshold can be a preconfigured value stored by each user equipment UE1 to UE12. For the slice demand loads p1, p2, p3, ..., p NBased on the value of, user equipments UE1 to UE12 can also obtain the required network slice and the values of service quality parameters required in the slice request to be permitted by the slice controller SC. User equipments UE1 to UE12 can also induce each user request into a plurality of demands to allocate resources to various network slices. Therefore, the slice demand load p of a network slice s with a large (i.e., exceeding a predetermined load threshold) value N Based on the value of, each user request can also be induced into a plurality of demands to allocate resources to various network slices. Therefore, the unique slice request of a network slice s with a large (i.e., exceeding a predetermined load threshold) value of slice demand load p s is replaced by some slice requests of some network slices with a small (i.e., smaller than a predetermined load threshold) value of slice demand load. Therefore, the proposed slice demand load method enables the formation of incoming user requests to suppress slice function stops and increase the admission of slice requests, increasing the reward and utility gain of system SYS. Therefore, such request formation depends on notifying user equipments UE1 to UE12 that send user requests of slice demand loads p1, p2, p3,..., p N in all network slices. In step S70, the value of load parameter λ and / or the values of slice service variables t1, t2, t3,..., t N can also be sent to user equipments UE1 to UE12. Such slice demand loads p1, p2, p3,..., p N depend on the derivative of slice utility functions U1, U2, U3,..., U N approximated, estimated, and / or predicted by a machine learning model such as a deep reinforcement learning model equipped with a neural network. Therefore, an operation performed by the slice controller can be generated using a unique learning algorithm, for example, a unique neural network, while maximizing slice utility gains U1, U2, U3,..., U N .

[0148] In step S7, based on the utility gain or utility loss obtained in step S3, the slice controller SC determines the admission of the processed user request, that is, the utility gain of the network slice s.

Number

Number

Number

[0149] The decision made in step S7 affects the overall state of the network system SYS and generates a reward (either positive or negative) for the utility of the system SYS in step SS8. Such a reward can be learned by the deep reinforcement learning model DQL, for example, as the utility value of the overall utility function U of the network system SYS, corresponding to the state of the system observed in step S2, the utility gain obtained in step S3, and the decision made by the slice controller SC in step S7.

[0150] In one embodiment, the update step S60 of the value of the load parameter λ of the following formula can also be executed after step S50.

[0151] Here,

Number

[0152] The gradient parameter δ k can be set and / or fixed, for example, to 10 -1 via a gradient descent technique. The gradient parameter δ k can also be a parameter to be updated, whose value decreases, for example,

Number

[0153] The current value λ of the load parameter λ k can be supplied to the deep reinforcement learning model DQL and can be memorized by the slice controller SC so as to be used in step S4 for future iterations.

[0154] In another embodiment, the load parameter λ can also be set as a fixed parameter.

[0155] In the present disclosure, the deep reinforcement learning model DQL learns the slice demand load p of all network slices s as the following formula.

Number

[0156] Therefore, the deep reinforcement learning model DQL can learn the slice utility function U s but can also learn the slice utility gain ∂U s / ∂t s as well.

[0157] Next, refer to FIG. 4. FIG. 4 shows the steps of executing a slice admission control mechanism according to an embodiment of the present disclosure. In such an embodiment of the present disclosure, the slice controller SC optimizes the slice admission control mechanism by leveraging the flexibility in the specifications of several applications and services executed by the application layer APPC.

[0158] The steps of the slice admission control mechanism presented in FIG. 3 are the same as those in FIG. 4 except for block step B1. In an embodiment of the present disclosure, block step B2 can be executed by the slice controller SC as detailed in FIG. 4.

[0159] In block step B2, step S4 is the same as that in block step B1.

[0160] In step S51, the utility gain

Number

Number

Number

[0161] In step SS61, the slice controller SC receives a response signal from the application layer APPC in response to the gain check signal. Such a response signal can include a controller signal indicating that relaxation of the constraint function

Equation

[0162] In step S61, upon receiving the response signal in step SS61, the slice controller SC then moves on to update the value of the budget parameter I as follows to obtain the increased optimal gain of the utility function U s

Equation

Equation

[0163] On the other hand, if the utility gain obtained by relaxing the constraint function has not changed significantly (i.e., increased or decreased), step SS61 is not performed, or indicates rejection of constraint relaxation. A significant change in the utility gain can be understood, for example, in terms of absolute value, as the utility gain exceeding a preconfigured gain threshold. Such a gain threshold is set, for example, to 5 percent. In such a case, step S61 is not executed by the slice controller SC (or, equivalently, the updated value of the resource budget parameter I is the current value of the resource budget parameter, i.e., I k+1 = I k is selected). Such an update of the resource budget parameter I is performed when the resource budget parameter I is less than the maximum resource budget parameter related to the size of the resource pool of the network system SYS, for example. The service level specification (SLS) managed by the application layer is strict, and / or there is no flexibility in the budget parameter I, so the application layer APPC can also reject constraint relaxation. In another embodiment, the updated value of the resource budget parameter I can be calculated by the application layer APPC, and step S61 consists of receiving such an updated value transmitted in step SS61.

[0164] By means of such a block step B2, the flexibility of the application layer APPC can be utilized to further optimize the utility gain obtained by permitting the processed slice request.

[0165] In step S7, based on such an increased optimal gain, the slice controller SC then determines whether to permit the incoming user request. On the other hand, if the obtained utility gain (regardless of whether the constraint function is relaxed) is negative (i.e., there is a loss of utility), the incoming user request is rejected.

[0166] Similar to the embodiment presented in FIG. 3, the update step S60 of the value of the load parameter λ can be executed after step S4. Such an update of the load parameter λ can improve the resource efficiency of the network system SYS at the application layer APP level.

[0167] Block steps B1 and B2 are represented in FIGS. 3 and 4 respectively. For clarity, block steps B1 and B2 are detailed separately in two different figures, but in the preferred embodiment of the present disclosure, block steps B1 and B2 are executed in combination, cumulatively, continuously and / or simultaneously between step S3 and step S7 as shown in FIGS. 3 and 4. In fact, the proposed slice admission control mechanism can simultaneously depend on the following. To form future user requests received by the slice controller SC, the slice demand load p of user equipment UE1 to UE12 in the deployment area DA s is provided to form user requests. Such a formation mode is represented in block step B1. Relaxing constraints to further increase the utility gain U of the network slice s by utilizing the potential flexibility of the application layer APPC. Therefore, the resource budget I of the network system SYS can be dynamically updated through the application layer APPC. Such a constraint relaxation mode is represented in block step B2. s The proposed slice admission control mechanism described in FIGS. 3 and / or 4 can be executed as a continuous loop, for example, until all available resources of the system SYS are allocated to all network slices.

[0168]

[0169] ​Next, refer to FIG. 5. FIG. 5 schematically shows the optimization of a slice admission control mechanism executed by a slice controller SC within a 5G communication network system SYS for the present disclosure.

[0170] As represented by arrow B1 in FIG. 5 * the slice controller SC can transmit and update slice demand loads p1, p2, p3, ..., p N through the 5G core network in each iteration of the proposed slice admission control mechanism. Such slice demand loads p1, p2, p3, ..., p N are then transmitted to user equipment UE1 to UE12 in a deployment area DA, for example, via a shared access network SAN. The slice controller SC, as represented by arrow B2 in FIG. 5 * can also notify the application layer APPC of an increase in gain that may occur in the utility function U s due to a change in the budget parameter I in each iteration of the slice admission control mechanism. Thus, the application layer APPC may relax the service level specification (SLS) managed at the edge computing level EC to optimize the utility gain at such a stage.

[0171] Next, refer to FIG. 6. FIG. 6 schematizes the structural architecture of a slice controller SC that executes a slice admission control mechanism within a 5G communication network system SYS.

[0172] The slice controller SC includes an input module INP that enables the slice controller SC to receive data and control signals from the control planes of the 5G core network CNW and the application layer APPC. The input module INP receives a slice request for resource allocation to be processed and the observed state of the network system SYS provided by the core network CNW.

[0173] The slice controller SC can execute a plurality of iterations of a slice admission control mechanism to manage the admission of user requests for slice resources in the network system SYS. To this end, the slice controller SC includes a processing unit PROC that executes the step of executing the slice admission control mechanism. In the present disclosure, such a processing unit includes the following. Slice demand load module SDL. By this module, the slice controller SC can obtain and update the values of slice demands p1, p2, p3,..., p N and communicate with the core network CNW to send such values to the user devices UE1 to UE12, for example, via the output module OUT of the slice controller SC. Load parameter update module LPU. By this module, the slice controller SC can quantify the cost of relaxing the constraint function when searching for an optimized utility gain by obtaining and updating the load parameter λ. Such a load parameter update module LPU also communicates with the core network CNW to send such updated values of the load parameter λ to the user devices UE1 to UE12, for example, via the output module OUT of the slice controller SC. Resource budget update module RBU. By this module, the slice controller SC can obtain the updated value of the budget parameter I set by the application layer APPC to obtain the optimized utility gain. Such a resource budget update module RBU communicates with the edge computing level EC, more specifically, the control plane of the application layer APPC, to notify the application layer APPC of the potential utility gain of relaxing the constraint function (i.e., changing the value of the budget parameter I), for example, via the output module OUT of the slice controller SC. The machine learning module DRL. With this module, the slice controller can determine the optimal slice admission policy through learning in successive iterations of the slice admission control mechanism. Therefore, the machine learning module DRL can execute a Q-learning model or a deep reinforcement learning model to learn and replay regarding the states, actions, and subsequent utility gains or losses of previous iterations of the slice admission mechanism. The machine learning module DRL can receive the supply of data from other modules of the slice controller SC such as the load parameter update module LPU and the slice demand load module SDL, and other data received and / or acquired by the slice controller SC.

[0174] The slice controller SC also includes a storage memory MEM. Such a storage memory MEM can also include a volatile memory. The storage memory MEM can store, for example, the values of the slice demands loads p1, p2, p3, ..., p N and the values of the slice service variables t1, t2, t3, ..., t N and the value of the load parameter λ and the value of the budget parameter I.

[0175] Next, refer to FIG. 7. FIG. 7 shows a flowchart of the steps executed by the user equipment to send slice requests to the communication network system SYS according to the requested method proposed in the present disclosure.

[0176] In step S100, a user equipment (for example, any user equipment among UE1 to UE12) deployed within the deployment area DA of the network system SYS can receive the (current) values of the slice demand loads p1, p2, p3, ..., p N associated with the plurality of network slices deployed by the network system SYS respectively.

[0177] Such slice demand loads p1, p2, p3, ..., pN reflects the impact on resources of ongoing slice requests received by the network system SYS for each network slice. Such slice demand loads p1, p2, p3, ..., p N reflects the resource consumption of incoming slice requests for each network slice. In other words, the greater the slice demand loads p1, p2, p3, ..., p N the greater the resource consumption of the demand parameters of the slice requests for such network slices.

[0178] In an optional step S200, the user equipment can also receive the (current) value of the load parameter λ associated with the current network resource state in the entire network system SYS. Based on such a current value of the load parameter λ and the received slice demand loads p1, p2, p3, ..., p N and, in an optional step S300, the user equipment can determine the slice utility gain

Equation

Equation

[0179] Therefore, the user equipment can locally estimate the service utility for each network slice based on the ongoing incoming slice requests.

[0180] Thus, in step S400, at least the received slice demand loads p1, p2, p3, ..., p NOr, based on the slice utility gain obtained in the optional step S300, the user equipment determines the slice request parameters associated with each of the slice requests of the user equipment to be transmitted to the network system SYS. Such slice request parameters can be, for example, the service performance requirements of the transmission service executed by the user equipment. Such slice request parameters can include, for example, a delay tolerance value, a required throughput value, a required reliability level, etc. At least the received slice demand loads p1, p2, p3,..., p N Based on this, the user equipment can adapt the slice request parameters to different network slices. The user equipment can determine the slice request parameters so as to reduce the resource consumption of the demand parameters on the network slice as much as possible. In other words, the user equipment can adapt (or form) the slice request parameters so as to have the slice requests of one or more network slices with the lowest slice demand loads p1, p2, p3,..., p N

[0181] In step S500, the user equipment can transmit a slice request (which can be considered as a future slice request for the network system) to the network system, for example, the buffering unit BF or the slice controller SC of the network system SYS. Such a slice request can include the adapted slice request parameters formed in step S400.

[0182] ​In step S600, the user equipment can potentially receive data related to the decision of admission of slice requests. For example, the user equipment can receive one or more network resources of the slice requests. This means that such slice requests are permitted by the slice controller. Step S600 can also include the expiration of a timer without receiving network resources from the network system SYS. This means that the slice requests are continuously buffered (therefore not permitted by the slice controller).

[0183] At least the received slice demand loads p1, p2, p3,..., p N By adapting the slice request parameters in step S400 based on, the user equipment maximizes the opportunity to permit the slice requests in step S600.

[0184] Such a request method can be executed, for example, by a user processing circuit (not shown) of the user equipment. Such a user processing circuit includes, for example, at least a user processing unit (or user processor).

Claims

1. An admission control method executed by a slice controller that performs admission control of slice admission requests in a communication network system, wherein the communication network system deploys a plurality of network slices, The admission control method includes: a) receiving at least one slice admission request from at least one user equipment deployed in the communication network system, the slice admission request including data related to slice request parameters; b) obtaining network parameters related to the overall state of network resources in the plurality of network slices; c) For each of the plurality of network slices, a slice service variable related to the network resources required from the network slice to permit the slice admission request, the slice service variable being determined based on at least the slice request parameters and the network parameters; a slice utility gain related to the change in the slice utility function expected from the utilization of the network resources required from the network slice to permit the slice admission request, the slice utility function being dependent on the slice service variable associated with each of the plurality of network slices, and the slice utility gain being determined based on at least a machine learning model; a slice demand load related to the network resources consumed in the network slice when the slice admission request is permitted, the slice demand load being determined based on at least the slice utility gain of the network slice; and calculating; d) transmitting the slice demand load associated with each of the plurality of network slices to the user equipment; e) supplying the slice demand load associated with each of the plurality of network slices to the machine learning model; f) determining, based on the slice utility gain, a decision regarding the permission of the slice admission request; An admission control method including

2. The admission control includes performing at least a plurality of repetitions of the steps a) to f), and two consecutive repetitions k and k + 1 are spaced based on a repetition parameter. The admission control method according to claim 1.

3. The slice utility gain is obtained based on learning of at least the slice demand load supplied to the machine learning model in the previous iteration k in the current iteration k + 1. The admission control method according to claim 2.

4. The slice demand load is obtained as the following formula, 【Number 1】 where p ｓ is the slice demand load of the network slice, λ is a load parameter, and the load parameter is related to the total amount of network resources in the plurality of network slices when at least one slice admission request from the at least one user equipment is permitted. t １ ,..., t ｓ ,..., t Ｎ is the slice service variable associated with each of the plurality of network slices, 【Number 2】 is the slice utility gain of the network slice. The admission control method according to any one of claims 1 to 3.

5. Before obtaining the slice demand load, obtaining a value of the load parameter, further including The value of the load parameter is updated as the following formula between two consecutive iterations k and k + 1. 【Number 3】 where λ ｋ＋１ is the updated value of the load parameter in the iteration k + 1, λ ｋ is the value of the load parameter in the iteration k, I is a resource budget parameter, p ｓ is the slice demand load of the network slice, t ｓ is the slice service variable of the network slice, 【Number 4】 is the difference obtained in the iteration k. δ ｋ is the gradient parameter calculated in the iteration k, and the gradient parameter is related to the rate of change of the load parameter with respect to the iteration k and the iteration k+1. The admission control method according to claim 4, combined with claim 2.

6. supplying the load parameter to the machine learning model, further including The slice utility gain is obtained based on learning of at least the load parameter supplied to the machine learning model in the previous iteration k in the current iteration k + 1. The admission control method according to claim 4 combined with claim 2.

7. transmitting the load parameter to at least the user equipment, further including. The admission control method according to claim 4.

8. After a predetermined number of repetitions, the admission control is performed based on a sub - part of the plurality of network slices, and the sub - part is composed of network slices in which the corresponding slice demand load is smaller than a predetermined load threshold. The admission control method according to claim 2.

9. The decision to permit the slice admission request is based on optimizing the service utility function of the plurality of network slices, and the service utility function is represented by the following formula: 【Number 5】 The service utility function is constrained by a constraint function represented by the following formula: 【Number 6】 Here, U ｓ (t １ , t ２ ,..., t Ｎ ) is the service utility function of the network slice, s is an index that identifies one network slice among the plurality of network slices, N is the number of network slices in the communication network system, p ｓ is the slice demand load of the network slice, t ｓ is the slice service variable of the network slice, I is a resource budget parameter set by the application layer of the communication network system, and the resource budget parameter I is related to the total amount of network resources allocated to the plurality of network slices configured by the application layer. The admission control method according to any one of claims 1 to 3.

10. Before the step d), Transmitting a gain check signal to the application layer, Further comprising, The gain check signal includes data related to a change in at least one slice utility gain considering the updated value of the resource budget parameter. The admission control method according to claim 9.

11. Based on the response signal received from the application layer, obtaining an updated value of the resource budget parameter by the following formula between two consecutive iterations k and k + 1: 【Number 7】 Further comprising, here, I ｋ＋１ is the updated value of the resource budget parameter in the iteration k + 1, I ｋ is the value of the resource budget parameter in the iteration k, λ ｋ is the value of the load parameter in the iteration k, β ｋ The admission control method according to claim 10 combined with claim 2, wherein β is the gradient ascent parameter calculated in the iteration k.

12. A slice controller configured to execute a slice admission control mechanism in a communication network system, the communication network system deploys a plurality of network slices, and the slice controller includes at least A processing unit, and A non-transitory computer-readable medium including stored instructions that configure the slice controller to execute the admission control method according to any one of claims 1 to 3 when executed by the processing unit, A slice controller comprising.

13. An admission request method executed by a user equipment for sending a slice admission request for at least one selected network slice in a communication network system that deploys a plurality of network slices, The slice admission request includes data related to slice request parameters, and the admission request method receiving a slice demand load associated with each of the plurality of network slices, each said slice demand load being related to the network resources consumed in the corresponding network slice when the slice admission request is permitted; adapting the slice request parameters of the slice admission request to request network resources for at least one selected network slice among the plurality of network slices based on at least the received slice demand load, the selected network slice having a corresponding slice demand load less than a predetermined load threshold; sending the slice admission request based on the adapted slice request parameters; An admission request method comprising:

14. A user equipment configured to send a slice admission request for at least one selected network slice in a communication network system deploying a plurality of network slices, a user processing unit; a non-transitory computer-readable medium for the user storing instructions which, when executed by the user processing unit, configure the user equipment to execute the admission request method according to claim 13; A user equipment comprising:

15. An admission control method according to any one of claims 1 to 3, an admission request method according to claim 13, A computer program product comprising program instruction code stored in a computer-readable medium that executes one of them.

Images