Open radio access network slice orchestration method and related device

Abstract

A method for slice orchestration of open radio access network, comprising: receiving a slice service request; generating a structured constraint object based on the slice service request; obtaining a system state set; determining a minimum continuous central processing unit (CPU) resource allocation condition that meets a slice service level agreement requirement; and determining a slice orchestration result based on the structured constraint object, the system state set, and the minimum continuous CPU resource allocation condition; wherein the slice orchestration result comprises a user association relationship, an open distributed unit deployment location, an open centralized unit deployment location, and a CPU resource allocation scheme. Accordingly, an electronic device, a computer readable storage medium, and a computer program product are further disclosed.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to an open wireless access network slicing and orchestration method and related apparatus. Background Technology

[0002] With the continuous evolution of fifth-generation mobile communication networks (5G) and future network architectures, Open Radio Access Network (O-RAN) has gradually become an important implementation method for new radio access networks due to its features such as hardware and software decoupling, open interfaces, and functional virtualization. Compared with the traditional closed radio access network architecture, O-RAN decouples baseband processing functions into multiple functional entities such as Open Centralized Unit (O-CU), Open Distributed Unit (O-DU), and Open Radio Unit (O-RU), and achieves interconnection and interoperability between these functional entities through standardized interfaces. This allows radio access network functions to be flexibly deployed on multi-layer cloud infrastructures such as edge cloud and regional cloud in the form of virtualized network functions or cloud-native network functions.

[0003] Under the O-RAN architecture, network slicing technology can provide differentiated service guarantees for different service types, such as enhanced mobile broadband, ultra-reliable low-latency communication, and massive machine-type communication. Network slicing (NS) is a core technology of 5G and next-generation mobile communication networks. Through technologies such as virtualization, Software Defined Networking (SDN), and Network Function Virtualization (NFV), it logically divides the same physical communication infrastructure (including base stations, transmission optical cables, core network equipment, servers, etc.) into multiple isolated, independently operated, and customized virtual private networks. To meet the heterogeneous requirements of different slices for latency, reliability, bandwidth, and resource isolation, operators typically need to dynamically make multiple orchestration decisions based on network load status, wireless coverage conditions, and cloud platform resources, including user association, O-DU / O-CU function deployment, computing power quota allocation, and wireless resource scheduling. Therefore, how to achieve efficient, reliable and executable resource orchestration in a complex environment where multiple cloud layers and multiple slices coexist has become an important technical problem in the O-RAN field. Summary of the Invention

[0004] In view of this, an open wireless access network slicing and orchestration method, an electronic device, a computer-readable storage medium, and a computer program product are provided to solve or partially solve the above problems.

[0005] In one scenario, the aforementioned open radio access network (RAN) slicing orchestration method may include: receiving a slice service request; generating a structured constraint object based on the slice service request; obtaining a system state set; determining the minimum consecutive CPU resource allocation conditions that satisfy the slice service level agreement requirements; and determining a slice orchestration result based on the structured constraint object, the system state set, and the minimum consecutive CPU resource allocation conditions; wherein the slice orchestration result includes: user association relationships, open distributed unit deployment locations, open centralized unit deployment locations, and CPU resource allocation schemes.

[0006] In addition, the aforementioned electronic device may include: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned open radio access network slicing orchestration method.

[0007] The aforementioned non-transitory computer-readable storage medium stores computer instructions that are used to cause a computer to execute the aforementioned open radio access network slicing and orchestration method.

[0008] The aforementioned computer program product includes computer program instructions that, when executed on a computer, cause the computer to perform the aforementioned open radio access network slicing and orchestration method.

[0009] Therefore, it can be seen that the aforementioned open radio access network slicing orchestration method, electronic devices, computer-readable storage media, and computer program products can be centered on an orchestration agent deployed on a Non-RT RIC. This integrates large language model-driven service intent parsing, rule constraint generation, burst traffic effective bandwidth modeling, cloud platform dynamic computation jitter modeling, probabilistic latency guarantee solution based on random network calculus, joint deployment and computing power allocation optimization, and O-RAN standard interface execution and feedback updates into a unified closed-loop orchestration mechanism. This mechanism can support operators in inputting slice service requirements through natural language or structured templates, and can also convert these requirements into verifiable mathematical constraints that meet latency and default probability requirements, further mapping them to deterministic orchestration results for O-DU, O-CU deployment, and CPU quota configuration. By introducing basic computation jitter measurement and load-related dynamic jitter modeling, this application can more accurately characterize computing power fluctuations in multi-tenant cloud environments, avoiding SLA distortion caused by existing fixed service rate assumptions. By deriving the minimum continuous CPU resource limit based on effective bandwidth and probabilistic latency boundaries, this application can reduce the computing power waste caused by peak reservation while ensuring slice latency and reliability requirements. By mapping the solution results to the O2, O1, A1, and E2 interfaces for execution, this application can achieve end-to-end closed-loop control of cloud-side instance orchestration, network element configuration, and radio-side scheduling, and can continuously update model parameters and re-orchestration strategies based on real-time operational feedback. Based on this, the slice orchestration scheme presented in this application has strong interpretability, verifiability, engineering feasibility, and system scalability, and is particularly suitable for O-RAN network slicing scenarios with strict latency and reliability requirements in multi-layer cloud environments. Attached Figure Description

[0010] To more clearly illustrate the technical solutions, the accompanying drawings used in the examples or related technical descriptions will be briefly introduced below. Obviously, the accompanying drawings described below are merely examples, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0011] Figure 1 This illustrates an application scenario example of the open radio access network slicing orchestration system described in some embodiments.

[0012] Figure 2 This shows an example of the implementation flow of the open radio access network slicing orchestration method described in some embodiments.

[0013] Figure 3 The implementation flow for determining the minimum contiguous CPU resource allocation conditions that meet the requirements of the slice service level agreement is shown in some embodiments.

[0014] Figure 4 The document provides a detailed process for determining the slice arrangement result using a hierarchical heuristic optimization method as described in some embodiments.

[0015] Figure 5 The implementation flow of executing slice orchestration results as described in some embodiments is shown.

[0016] Figure 6 The implementation flow of the feedback monitoring and dynamic reordering methods described in some embodiments is shown.

[0017] Figure 7 This diagram shows the internal structure of an open wireless access network slicing and orchestration apparatus as described in some embodiments.

[0018] Figure 8 An example of an electronic device hardware architecture for implementing an open radio access network slicing orchestration method is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages clearer, the technical solutions will be further explained in detail below with reference to specific examples and accompanying drawings.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used should have the common meaning understood by a person with general skills in the field. The terms "first," "second," and similar words used in the examples do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Words such as "including" or "contains" mean that the element or object preceding the word covers the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; these relative positions may change accordingly when the absolute position of the described object changes.

[0021] It is understandable that before using the technical solutions in each example, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0022] For example, upon receiving a user's proactive request, a prompt message can be sent to the user, explicitly informing them that the requested operation will require the acquisition and use of their personal information. This allows the user to choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the technical solution.

[0023] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0024] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the described implementation method. Other methods that comply with relevant laws and regulations may also be applied to the described implementation method.

[0025] The term "in response to" as used herein refers to a state in which a corresponding event occurs or a condition is satisfied. It will be understood that the timing of subsequent actions performed in response to such event or condition is not necessarily strongly correlated with the time when the event occurs or the condition is met. For example, in some cases, subsequent actions may be performed immediately upon the occurrence of the event or the fulfillment of the condition; while in others, they may be performed some time after the occurrence of the event or the fulfillment of the condition.

[0026] As mentioned earlier, how to achieve efficient, reliable and executable resource orchestration in a complex environment where multiple cloud layers and multiple slices coexist has become an important technical problem in the O-RAN field.

[0027] Currently, most research on O-RAN network element deployment and resource allocation employs static rules, average traffic models, or deterministic latency models for analysis and decision-making. For example, some methods reserve and allocate O-DU and O-CU deployment locations and server central processing unit (CPU) resources based on fixed bandwidth requirements, average service rates, or estimated peak loads. While these methods are relatively simple to implement, they struggle to accurately reflect the bursty, random, and time-varying characteristics of real mobile network services. Especially in enhanced mobile broadband and low-latency, high-reliability service scenarios, service traffic often exhibits significant burst arrival behavior. If resource allocation is based solely on average values, it can easily lead to increased queuing latency and service breaches during short-term bursts. However, reserving resources based on peak traffic for extended periods results in a significant amount of computing resources being idle, reducing the efficiency of cloud platform resource utilization.

[0028] In addition to the randomness of service traffic itself, the cloud computing infrastructure upon which O-RAN relies also commonly suffers from fluctuations in processing capacity. In actual deployments, O-DU and O-CU often run on general-purpose servers, virtual machines, or container environments. Their processing performance is affected by various factors such as multi-tenant contention, virtualization overhead, operating system scheduling, interrupt activity, cache contention, memory access, and hardware heterogeneity, making it difficult to maintain constant stability. Especially in edge cloud environments, due to limited resource scale, rapid load changes, and high degree of sharing, server processing capacity often exhibits time-varying and random jitter characteristics. This type of computing jitter directly affects the processing latency and queue waiting latency of the protocol stack, making it difficult for deployment and allocation methods based on the assumption of a fixed service rate to continuously meet the latency and reliability requirements of strictly sliced ​​services.

[0029] On the other hand, with the development of large-scale models and intelligent agent technologies, some solutions have begun to introduce natural language interaction, automatic policy generation, or intelligent recommendations into the field of network orchestration, enabling operators to trigger network configuration through business descriptions or policy intentions. However, most existing solutions remain at the level of policy suggestion generation or text-to-configuration template conversion, often lacking the ability to map high-level business intentions to strict mathematical constraints and verifiable resource boundaries. The output results usually lack clear reliability guarantees and are difficult to directly apply to O-RAN slicing scenarios with strict requirements for probabilistic latency, processing stability, and resource isolation. Especially in actual engineering systems, simply generating deployment results from large models may also lead to semantic ambiguity, parameter out-of-bounds errors, constraint conflicts, and non-executability issues, making it difficult to meet the O-RAN standard interface control and live network security operation requirements.

[0030] Furthermore, existing O-RAN orchestration technologies generally suffer from insufficient coupling between upper-layer requirements, lower-layer execution, and operational feedback. Some solutions only calculate network element deployment results in offline environments, lacking an execution flow that integrates with standard interfaces such as the Operation and Maintenance Management Interface (O1), Service Orchestration / Automation Interface (O2), Wireless Intelligent Management Interface (A1), and Open Fronthaul Intelligent Interface (E2). While some solutions achieve cloud-side instantiation or wireless-side scheduling, they lack a unified closed-loop mechanism encompassing service intent parsing, constraint generation, deployment solving, policy distribution, operational monitoring, and dynamic re-orchestration. Therefore, existing technologies struggle to simultaneously achieve the following objectives: supporting service intent input described by natural language or policy templates, transforming these intents into verifiable computing power and deployment constraints that meet latency and reliability requirements, and further forming a closed-loop control through O-RAN standard interface execution and feedback updates.

[0031] Therefore, a new O-RAN slicing orchestration technology solution is urgently needed. After receiving service intent or slice request, it can use intelligent agents to complete intent understanding and constraint generation, establish a probabilistic latency guarantee model by combining burst traffic characteristics and cloud computing jitter characteristics, convert the service level agreement (SLA) of the slice into deterministic network element deployment and computing power configuration conditions, and complete the actual orchestration execution and dynamic closed-loop optimization for O-RAN multi-layer cloud environment through interfaces such as O1, O2, A1, and E2.

[0032] Figure 1 This illustration shows an application scenario example of an O-RAN slicing orchestration system provided by an embodiment of this disclosure. For example... Figure 1As shown, in the above application scenario, the Business & Operation Support System (BOSS) 110 can send slice intents, user needs, policy preferences, and other information in natural language or slice service requests in structured templates to the Service Management and Orchestration (SMO) 120. The Non-Real-Time RAN Intelligent Controller (Non-RT RIC) 122 within the SMO 120 is the core processing unit. It can parse service intents, generate structured constraint objects, and solve probabilistic latency through the orchestration agent 124 deployed within it (also known as the Deterministic Slice Orchestration Agent). Based on this, it completes joint orchestration optimization and policy generation, and finally executes the policy and provides updates through a standard interface. Specifically, the orchestration agent 124 can include a service parsing module, an orchestration optimization module, and an execution module. The aforementioned business parsing module can be used to parse business intents, generate structured constraint objects, and solve probabilistic latency; the aforementioned orchestration optimization module can be used to complete joint orchestration optimization and policy generation; and the aforementioned execution module can be used to execute policies through standard interfaces and provide feedback updates. Furthermore, the service management and orchestrator 120 can issue user association policies to the Near-Real-Time RAN Intelligent Controller (Near-RT RIC) 130 via the A1 interface, which will then complete the radio resource allocation. The service management and orchestrator 120 can also interact with the O-RAN network element layer 140 via the O1 interface to collect performance data and issue network element parameter policies. The service management and orchestrator 120 can also collaborate with the Open Cloud (O-Cloud) platform 150 via the O2 interface to collect and calculate jitter characteristics and issue cloud-native orchestration commands. Furthermore, the various modules can achieve data interoperability and resource linkage through the E2 interface and instance mapping. As can be seen, in the above application scenarios, it can be used for slice services in O-RAN multi-layer cloud environment to perform business intent understanding, probability latency constraint generation, network element placement decision, computing resource configuration, standard interface distribution and operation feedback update.

[0033] Figure 2This illustration shows an O-RAN slicing orchestration method provided by an embodiment of the present disclosure. The O-RAN slicing orchestration method described above can be executed by an orchestration agent 124 deployed within a non-real-time radio access network intelligent controller 122 within a service management and orchestrator 120. For example... Figure 2 As shown, the O-RAN slice orchestration method described above may include the following steps.

[0034] In step 210, a slice service request is received.

[0035] In step 220, a structured constraint object is generated based on the slice business request.

[0036] In step 230, the system state set is obtained.

[0037] In step 240, the minimum contiguous CPU resource allocation conditions that meet the requirements of the slice service level agreement are determined.

[0038] In step 250, the slice arrangement result is determined based on the structured constraint object, the system state set, and the minimum continuous CPU resource allocation condition.

[0039] In some cases, the above O-RAN slice orchestration method may further include: step 260, executing the slice orchestration result. Furthermore, in other cases, the above O-RAN slice orchestration method may further include step 270, feedback monitoring and dynamic re-orchestration.

[0040] The specific process of the above O-RAN slice arrangement method will be further explained in detail below with reference to the accompanying drawings and examples.

[0041] Regarding step 210 above, at the beginning of each orchestration cycle, the non-real-time radio access network intelligent controller within the service management and orchestrator 120 can receive slice service requests initiated by the upper-layer service operation support system. In some cases, the slice service requests can be in natural language form; or, the slice service requests can be in the form of predefined structured templates. For example, a slice service request in natural language form could be: "Open low-latency, high-reliability slices for industrial control terminals, prioritize deployment in the edge cloud, and allow control-related services to remain at the edge and non-critical services to migrate to the regional cloud when edge resources are insufficient." Furthermore, the structured template can include fields such as slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preference, and migration strategy.

[0042] For step 220 above, the orchestration agent deployed in the non-real-time radio access network intelligent controller can be used to parse the above slice service request and generate a structured constraint object containing slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preference and migration strategy.

[0043] Specifically, in step 220 above, in some cases, for slice service requests in natural language form, a large language model can be used to perform semantic understanding and key information extraction on the natural language form of the slice service request, transforming it into the aforementioned structured constraint object. For example, an intent parsing unit is generated based on the structure of the large language model and the structured constraint object, and the intent parsing unit is used to perform the aforementioned semantic understanding and key information extraction on the slice service request, thereby generating a structured constraint object. As mentioned above, the aforementioned structured constraint object can at least include one or a combination of information such as slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preference, and migration strategy. In a specific example, the aforementioned structured constraint object can be specifically represented as: .in, It can represent the slice type. It can indicate the service area. It can represent the end-to-end latency budget. This represents the threshold for the probability of default due to time delay. It can indicate priority. It can express deployment preferences, and This can represent the migration strategy. Through the above processing, the transformation from business requirements to structured constraint objects can be completed, thus providing a prerequisite for subsequent mathematical solutions and standardized execution.

[0044] Furthermore, in other cases, since the aforementioned structured template can include fields such as slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preferences, and migration strategy, slice service requests based on structured templates can be parsed based on predefined structured templates. For example, the aforementioned fields such as slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preferences, and migration strategy can be directly extracted from the slice service request, and a structured constraint object can be generated based on the extracted fields. Structured constraint objects of a structure.

[0045] Furthermore, after generating the aforementioned structured constraint objects, the rule constraint generation unit can perform rule validation on these structured constraint objects based on pre-defined validation rules, thereby obtaining structured constraint objects that conform to the validation rules. In some cases, the validation rules may include: parameter range checking, field completion, and illegal request filtering. Regarding the parameter range checking, during rule validation, a pre-defined parameter range can be applied to check each parameter in the structured constraint object, and parameters that do not meet the parameter range requirements can be adjusted to conform to them. For example, when the latency budget generated by the large language model is lower than the system's supported lower limit or the default probability threshold exceeds a preset range, the rule constraint generation unit can automatically correct the parameters to the closest executable configuration or return an alarm request for manual confirmation. Regarding field completion, during rule validation, pre-defined default parameter values ​​can be used to complete the actual parameter values ​​in the slice business request. Regarding illegal request filtering, during rule validation, pre-defined filtering conditions can be applied to the structured constraint objects to remove those that do not meet the filtering conditions.

[0046] Regarding step 230 above, in some cases, the aforementioned system state set may include: O-RAN state information, cloud platform telemetry information, RIC information, and auxiliary information, or any combination thereof. In some specific examples, the aforementioned system state set can be uniformly represented as: ,in, It can represent O-RAN status information; It can represent telemetry information from the cloud platform; It can represent RIC information; and, It can represent auxiliary information. The above system state set can be used to drive business traffic modeling, dynamically calculate jitter modeling, and jointly optimize solutions.

[0047] The orchestration agent can obtain the aforementioned O-RAN status information through the O1 interface. Specifically, the O-RAN status information may include one or any combination of the following: O-RAN network element management status, network topology and radio resource occupancy status, radio coverage status, link reachability, and alarm information. The O-RAN network element management status may include the configuration status of O-RU, O-DU, and O-CU.

[0048] The orchestration agent can obtain the aforementioned cloud platform telemetry information through the O2 interface. Specifically, the cloud platform telemetry information may include one or any combination of the following: available CPU capacity, container instance running status, resource usage, node health, instance location, and processing latency sampling information.

[0049] The aforementioned RIC information can be obtained by the orchestration agent through Near-RT RIC or its associated monitoring module. Specifically, the RIC information may include one or any combination of the following: wireless load summary, user access status, handover status, and E2-side execution feedback.

[0050] In addition to the aforementioned auxiliary information, the orchestration agent can also collect other auxiliary information related to subsequent orchestration. This auxiliary information may include: the maximum allocatable CPU resources of the server. Wireless resource block capacity Information such as the connection relationship between the server and the O-RU, the user's current access location, and the deployment status of existing network elements, or any combination thereof.

[0051] In step 230 above, the O-RAN state information, cloud platform telemetry information, RIC information and auxiliary information can be combined to form the complete state vector of the current orchestration cycle.

[0052] Regarding step 240 above, in some cases, Figure 3 The implementation flow for determining the minimum contiguous CPU resource allocation conditions that meet the requirements of the slice service level protocol, as described in some embodiments, is shown. For example... Figure 3 As shown, the method for determining the minimum contiguous CPU resource allocation conditions that meet the requirements of the slice service level agreement may include the following steps.

[0053] In step 310, determine the server. Slice-oriented Example of the function Aggregated bandwidth.

[0054] In step 320, determine the server. Slice-oriented Dynamically calculate jitter parameters.

[0055] In step 330, the server is determined based on aggregated bandwidth and dynamically calculated jitter parameters. Slice-oriented Example of the function The minimum amount of consecutive CPU resources allocated.

[0056] In step 340, the minimum amount of consecutive CPU resources is used as the minimum consecutive CPU resource allocation condition.

[0057] The following will further explain in detail the specific process for determining the minimum consecutive CPU resource allocation conditions, with reference to the accompanying drawings and examples.

[0058] It should be noted that, in some embodiments of this disclosure, the aforementioned server may refer to a server that carries an O-DU function instance or a server that carries an O-CU function instance.

[0059] Regarding step 310 above, in some cases, the user service flow arrival process can be modeled using an On-Off Markov Modulated Poisson process. Specifically, the parameters of the On-Off Markov Modulated Poisson process model can be obtained through historical service statistics, online traffic observation, or sliding window estimation. The On-Off Markov Modulated Poisson process can describe the burst behavior of the service flow through two states, where the service arrives at a peak rate in the "On" state and at a low or near-zero arrival rate in the "Off" state. If we assume the state transition matrix is... The arrival rate matrix is Then the user The effective bandwidth can be expressed as .in, It can be a constraint severity parameter; This can represent the spectral radius. It can be understood that, through the above parameters, the pressure exerted by user services on server processing capacity and wireless resources can be described statistically.

[0060] Furthermore, in O-RAN function decomposition scenarios, the traffic volumes entering O-DU and O-CU are typically different. Therefore, a traffic scaling factor can be introduced to represent the effective bandwidth of services on the O-CU side as a function related to the effective bandwidth on the O-DU side. Specifically, by aggregating the effective bandwidth of users corresponding to the same slice function instance deployed on the same server, the aggregated bandwidth can be obtained. In other words, under the independence assumption, the orchestration agent targets the bandwidth deployed on the server... upper slice Functional Examples The server is obtained by summing the effective bandwidth of each user in the user set. Slice-oriented Example of the function aggregated bandwidth ,in, It can be represented as a server Slice-oriented Example of the function For example, deployed on a server O-DU or O-CU function instance on the device.

[0061] Regarding step 320 above, in this embodiment, to quantify the server's basic computational jitter, the orchestration agent samples the processing latency of protocol stack tasks within a preset statistical window to obtain the server's latency. Processing delay sample sequence within the above statistical window Furthermore, based on expressions Determine the server Basic calculation of jitter parameters .in, This represents the standard deviation calculation. The aforementioned processing latency samples can be obtained through O2 telemetry, container runtime statistics, kernel tracing, or protocol stack instrumentation. Furthermore, the actual processing capacity of a server is not only related to the number of allocated CPUs but is also affected by multi-tenant contention, system overhead, and load variations. Therefore, the orchestration agent performs basic jitter parameter calculations. Based on this, a dynamic jitter calculation model is constructed by introducing load-related terms. In one implementation, the server... Slice-oriented The dynamically calculated jitter parameters are expressed as follows: .in, It can represent a server The slice currently being carried Number of users; This can be a load sensitivity coefficient for cloud servers, used to characterize the jitter increment caused by an increase in the number of users. This load sensitivity coefficient can be obtained through offline fitting, for example, by using least squares fitting, linear regression, or online updating with a sliding window based on the standard deviation of processing latency samples under different numbers of users.

[0062] It is understandable that the above method can effectively establish the mathematical expression basis for the business arrival process and the server service fluctuation process, thereby providing input for the subsequent derivation of probabilistic latency guarantee constraints.

[0063] Regarding step 330 above, in some cases, the orchestration agent can use the effective bandwidth of the slice service obtained in the preceding process and the dynamically calculated jitter parameters to convert the slice service level agreement (SLA) requirements into server resource configuration conditions, that is, minimum continuous CPU resource allocation conditions.

[0064] More specifically, in step 330 above, firstly, the fronthaul delay and midhaul delay are determined based on the aforementioned system state set. For example, the fronthaul delay and midhaul delay can be calculated as deterministic transmission delays based on the positional relationships of the user-accessed O-RU, candidate O-DU, and candidate O-CU. Specifically, the positional relationships of the user-accessed O-RU, candidate O-DU, and candidate O-CU can be extracted from the network topology information in the O-RAN state information, and the fronthaul delay and midhaul delay can be calculated based on these positional relationships.

[0065] Furthermore, parameters such as end-to-end delay budget can be extracted from structured constraint objects.

[0066] Then, the total computational latency budget is determined based on the end-to-end latency budget, the fronthaul latency, and the midhaul latency. Specifically, if the fronthaul latency is... The latency of Communication University of China is ,slice The end-to-end latency budget is The total computational latency budget available for protocol processing and queuing can then be expressed as: In other words, the total computational delay budget mentioned above can be represented as a slice. The end-to-end latency budget is Compared to the prequel, the latency is The latency of Communication University of China is The difference between the sums.

[0067] Subsequently, the orchestration agent allocates the computational latency budget to the O-DU and O-CU processing stages according to a preset ratio or empirical rules, thereby obtaining the local computational latency budget for the O-DU functional instance. And local computational latency budget in O-CU functional instances As a unified expression, it can be used This refers to any of the functional instances on the aforementioned servers. Local computational delay budget.

[0068] In terms of service modeling, if the server slices Functions The allocated service rate Therefore, its service process can be described using a drift service model with dynamic jitter. Within the framework of stochastic network calculus, the steady-state queuing delay default probability can be represented by the following upper bound. .in, It can represent function On the server Slice-oriented Aggregated effective bandwidth; This can represent the corresponding dynamically calculated jitter parameters; It can be a random variable representing end-to-end processing latency; This can represent the end-to-end processing latency threshold, specifically the local computation latency budget. and The sum of these is the total computational delay budget mentioned above.

[0069] When the upper bound of the above-mentioned probability of default is not greater than the slice Given a default probability threshold (i.e., the default probability threshold in structured constraint objects) And the local delay budget satisfies When this is the case, the lower limit of continuous CPU resources, i.e., the minimum amount of continuous CPU resources, can be obtained by reverse calculation, and used as the minimum continuous CPU resource allocation condition mentioned above. In one embodiment, the minimum amount of continuous CPU resources is expressed as: .in, It can represent a server slices Functions The amount of continuous CPU resources allocated; This can improve the processing efficiency of the software RAN protocol stack; A fixed clock frequency can be set for the CPU; It can represent a functional instance. The local computational latency budget is calculated using this formula. Through this formula, the orchestration agent can explicitly translate abstract SLA requirements into lower bounds on CPU quotas on candidate servers, thereby achieving a mapping from probabilistic latency guarantees to deterministic computing power constraints.

[0070] Regarding step 250 above, the orchestration agent can jointly solve for the user-O-RU association strategy, O-DU / O-CU function deployment strategy, and CPU resource allocation strategy for each candidate server for different slice function instances based on structured constraint objects, system state set, and minimum continuous CPU resource allocation conditions, to obtain a deterministic slice orchestration result. In some cases, the above slice orchestration result may include: user association relationships, O-DU deployment locations, O-CU deployment locations, and CPU resource allocation schemes.

[0071] In some cases, the aforementioned joint solution can be achieved through iterative optimization. Specifically, the iterative optimization methods can include: heuristic search, metaheuristic optimization, mathematical programming, reinforcement learning, evolutionary search, or any combination thereof. In one specific implementation, a hybrid genetic algorithm guided by random network calculus based on an improved genetic algorithm can be used to perform joint iterative optimization of user association, open distributed unit deployment location, open centralized unit deployment location, and CPU resource allocation scheme.

[0072] In some cases, the orchestration agent can determine the slice orchestration results using a hierarchical heuristic optimization approach based on the aforementioned structured constraints, system state set, and minimum continuous CPU resource allocation conditions. As mentioned earlier, the slice orchestration results may include user relationships, open distributed unit deployment locations, open centralized unit deployment locations, and CPU resource allocation schemes. Figure 4 This paper provides a specific flowchart for determining the slice arrangement result using a hierarchical heuristic optimization approach. For example... Figure 4As shown, the above-mentioned hierarchical heuristic optimization method for determining the slice arrangement result may include the following multiple steps, and the above method can be executed by the arrangement agent.

[0073] In step 405, a set of candidate paths is generated for each user based on the network topology relationships in the system state set.

[0074] In some cases, the aforementioned candidate path set may include at least one candidate path. Each candidate path can be represented as a combination of user, open radio unit, open distributed unit candidate deployment server, and open centralized unit candidate deployment server. For example, for a user... The candidate paths generated for it can be represented as ,in, This indicates the open radio frequency units that users can access. This indicates a candidate open distributed unit deployment server. This indicates a candidate open centralized unit deployment server.

[0075] In step 410, the candidate path set is pre-pruned based on the wireless coverage relationship, link reachability relationship, server resource status and end-to-end latency budget in the system state set and the structured constraint object to obtain the feasible candidate path set for each user.

[0076] Specifically, in some cases, if a candidate path does not satisfy the wireless coverage relationship between the user and the open radio unit, the link reachability relationship between the open radio unit and the candidate open distributed unit deployment server, or the midhaul reachability relationship between the candidate open distributed unit deployment server and the candidate open centralized unit deployment server, or if the sum of its fronthaul latency and midhaul latency makes the remaining computational latency budget less than a preset threshold, then the candidate path is deleted. After pre-pruning, each user retains at least one feasible candidate path as the user's feasible candidate path set for subsequent orchestration optimization.

[0077] In step 415, an initial orchestration population is generated based on the set of feasible candidate paths for each user.

[0078] In some cases, each individual in the initial orchestration population is used to represent the path selection results for all users. In other words, an individual can consist of multiple gene bits, each corresponding to a user and recording the path index selected by that user from their set of feasible candidate paths.

[0079] In step 420, the individuals in the initial orchestration population are decoded to obtain the user-related variables, O-DU deployment variables, and O-CU deployment variables corresponding to each individual.

[0080] In step 425, the set of slice users carried on each server is determined based on user associations, O-DU deployment variables, and O-CU deployment variables.

[0081] Specifically, for the server Slice-oriented Example of the function ,in, It can be an open distributed unit function or an open centralized unit function. The orchestration agent can count the set of slice users processed by the server and determine the number of users the server is facing for that slice.

[0082] In step 430, CPU resources are allocated to the server for functional instances on the slice based on the slice user set.

[0083] Specifically, in some cases, step 460 above may include: First, the orchestration agent determines the aggregate bandwidth of the server for the functional instances on the slice and the dynamically calculated jitter parameters of the server for the slice based on the slice user set. Further, for each activated combination of server, slice, and functional instance, the orchestration agent performs inner-layer CPU resource allocation calculations to achieve CPU resource allocation of the server for the functional instances on the slice.

[0084] Specifically, for O-DU functional instances, the aggregated bandwidth can be obtained by summing the effective bandwidth of users deployed on the same server and belonging to the same slice; while for O-CU functional instances, the aggregated bandwidth can be further determined by combining the traffic scaling factor after functional decomposition. Meanwhile, the orchestration agent can also determine the server's dynamic jitter parameters for each slice based on the server's basic jitter parameters and the number of slice users currently supported by the server. It should be noted that the specific methods for determining the aggregated bandwidth and dynamically calculating jitter parameters can be found in the aforementioned examples and will not be repeated here.

[0085] Furthermore, the orchestration agent can perform a one-dimensional search within the preset range of constraint severity parameters; and under each candidate constraint severity parameter value, it calculates the candidate CPU resource allocation amount that meets the probability delay guarantee requirements based on aggregate bandwidth, dynamically calculated jitter parameters, local computation delay budget, and the delay default probability threshold of the slice; and selects the minimum value from the candidate CPU resource allocation amounts that meet the above minimum continuous CPU resource allocation conditions as the CPU resource allocation result for the server facing the slice function instance.

[0086] In step 435, constraint verification is performed on each individual in the initial orchestration population based on the CPU resource allocation results.

[0087] In some cases, the aforementioned constraint checks may include: user unique association constraint checks, wireless resource block capacity constraint checks, function unique deployment constraint checks, topology reachability constraint checks, server CPU capacity constraint checks, end-to-end latency constraint checks, and minimum contiguous CPU resource allocation condition checks. If an individual violates any of the above constraints, a constraint violation penalty can be set for that individual; if an individual satisfies all of the above constraints, then that individual is considered a feasible orchestration individual.

[0088] In step 440, the fitness value of each individual in the initial staging population is determined.

[0089] In some scenarios, the orchestration agent can calculate the current fitness value based on CPU resource allocation, server energy consumption coefficient, resource usage cost, deployment preference violation degree, migration cost, and constraint violation penalty. In one example, the fitness value can be obtained by weighting the server CPU resource allocation cost with the constraint violation penalty. For an individual that satisfies all constraints, its constraint violation penalty can be zero; for an individual that violates wireless resource, server capacity, or latency constraints, its constraint violation penalty can increase according to the degree of violation.

[0090] In step 445, the initial orchestration population is updated based on the fitness values ​​of each individual in the initial orchestration population.

[0091] Specifically, the aforementioned population update operations can include tournament selection, crossover, and mutation. The selection operation is used to retain the best-fit individuals for orchestration; the crossover operation is used to exchange the path selection results of some users among different individuals to generate new orchestration candidates; and the mutation operation is used to randomly change the candidate path indices of some users to expand the search space and avoid getting trapped in local optima.

[0092] Next, the orchestration agent can return to step 420 above and repeat steps 420-445, that is, repeatedly perform individual decoding, slice user set statistics, aggregate bandwidth calculation, dynamic jitter parameter calculation, CPU resource allocation calculation, constraint verification, fitness calculation, and population update operations until the iteration termination condition is met, and then continue to execute the subsequent step 450. Specifically, the above iteration termination condition may include one or any combination of the following: reaching the maximum number of iterations, fitness value convergence, and meeting the preset computation time limit.

[0093] In step 450, select the individual with the best fitness value and that meets the constraints from the updated orchestration population or the historical best individual as the target orchestration individual.

[0094] In step 455, a deterministic slice arrangement result is output based on the target arrangement individual.

[0095] As mentioned above, the slice orchestration results can include: user associations, O-DU deployment locations, O-CU deployment locations, and CPU resource allocation schemes. Therefore, the O-RAN slice orchestration method can jointly solve discrete orchestration decisions such as user association, open distributed unit deployment, and open centralized unit deployment with continuous resource decisions such as CPU resource allocation. Simultaneously, it reduces infeasible candidate paths through pre-pruning and calculates the minimum CPU resource allocation to meet probabilistic latency guarantees through inner-layer one-dimensional search, thereby reducing resource usage costs while meeting the requirements of the slice service level agreement.

[0096] Regarding step 260 above, the specific steps for executing the slice orchestration results may include: the orchestration agent maps the deterministic orchestration results obtained in step 250 to control signaling executable by the O-RAN standard interface and the cloud-native orchestration platform, completing the implementation from decision-making to execution. Specifically, step 260 above can be as follows: Figure 5 As shown, it includes the following steps.

[0097] In step 510, the O-DU deployment location, O-CU deployment location, and CPU resource allocation scheme are converted into O2 orchestration requests.

[0098] Specifically, the orchestration request may include one or any combination of parameters such as target server identifier, functional instance type, slice identifier, image identifier, number of instances, CPU quota parameters, and CPU cycle parameters. In one specific implementation, the orchestration request may be a set containing the following fields: .

[0099] In step 520, the O2 orchestration request is sent to the O-Cloud infrastructure platform via the O2 interface.

[0100] In step 530, the target O-DU and target O-CU are instantiated based on the O-DU deployment location and the O-CU deployment location.

[0101] In step 540, the CPU quotas for the target O-DU and the target O-CU are set.

[0102] Specifically, upon receiving an orchestration request, the O-Cloud infrastructure platform instantiates the target O-DU or target O-CU on the specified node. The O-Cloud infrastructure platform sets CPU quotas through container resource control mechanisms. For example, in a Kubernetes environment, the CPU usage limit for a container can be calculated based on CPU_Quota and CPU_Period, and resource limits can be implemented through underlying cgroup parameters.

[0103] In step 550, the initialization configuration is sent to the target O-DU and the target O-CU through the O1 interface.

[0104] In some cases, the above initialization configuration includes: network element identifier, network layer address, adjacent network element information, bearer connection parameters, slice instance identifier, service link configuration parameters, or any combination thereof.

[0105] Specifically, after the network element is instantiated, the orchestration agent can send initialization configurations to the newly instantiated or migrated O-DU and O-CU through the O1 interface. The O1 interface can use the NETCONF protocol in conjunction with the YANG data model to complete the configuration sending, thereby establishing the bearer relationship between O-DU and O-CU, and between O-DU and O-RU.

[0106] In step 560, A1 policy information is generated based on user association relationships.

[0107] In step 570, the A1 policy information is sent to the Near-RT RIC via the A1 interface.

[0108] Specifically, for wireless-side user association and policy enforcement, the orchestration agent can encapsulate user association results, slice priorities, and deployment preferences from structured constraint objects into A1 policy information and distribute it to Near-RTRIC. In some cases, the aforementioned A1 policy information can declaratively represent the mapping relationship, priority, and edge-first policy between the user group and the target O-RU of the target slice.

[0109] In step 580, scheduling guidance information is generated based on the A1 policy information.

[0110] In step 590, scheduling guidance information is sent to the target E2 node through the E2 interface.

[0111] Specifically, after receiving the A1 policy information, the Near-RT RIC can parse and generate near real-time control actions through its internal xApp, and then send scheduling guidance information to the lower-layer O-DU or related E2 nodes through the E2 interface to perform user access adjustment, PRB allocation priority control and slice bearer mapping operations.

[0112] In other words, the execution sequence of step 260 above can be summarized as follows: first, network element instantiation and computing power quota configuration are completed through the O2 interface; then, network element connection parameters and basic network configuration are distributed through the O1 interface; subsequently, higher-level policies are distributed to the Near-RT RIC through the A1 interface; and finally, the Near-RT RIC completes the wireless-side linkage execution via the E2 interface. This sequence ensures that cloud-side computing resources and network element bearer relationships are ready before implementing wireless-side access and scheduling policy switching.

[0113] Regarding step 270 above, after the orchestration execution is completed, the orchestration agent can enter the operation monitoring and dynamic re-orchestration phase. The implementation flow of the above feedback monitoring and dynamic re-orchestration method can be as follows: Figure 6 As shown, the specific steps may include the following.

[0114] In step 610, obtain slice arrangement feedback information.

[0115] Specifically, in some cases, the aforementioned slice orchestration feedback information may include one or any combination of O-RAN feedback information, cloud platform feedback information, and RIC feedback information. Based on this, step 510 may include one or a combination of the following steps: obtaining O-RAN feedback information through the O1 interface; wherein the O-RAN feedback information may include one or any combination of information such as network element alarms, connection status, and configuration change results; or obtaining cloud platform feedback information through the O2 interface; wherein the cloud platform feedback information may include one or any combination of information such as cloud platform instance running status, CPU utilization, resource contention, container throttling events, and processing latency statistics; or obtaining RIC feedback information through Near-RT RIC or other monitoring modules; wherein the RIC feedback information may include one or any combination of information such as radio resource utilization, user access changes, slice service quality indicators, and E2 execution results.

[0116] In step 620, it is determined whether to trigger re-arrangement based on the slice arrangement feedback information and the preset re-arrangement conditions.

[0117] In some cases, the O-RAN feedback information, cloud platform feedback information, and RIC feedback information can be compared with the current model parameters to update the basic and dynamic calculation jitter parameters, and to determine whether one or more pre-set re-arrangement conditions are triggered. Specifically, the re-arrangement conditions may include one or any combination of the following conditions: the actual processing latency of the slice service continuously approaches or exceeds the local budget threshold; the number of server CPU throttling events exceeds a preset threshold; the basic or dynamic calculation jitter parameters increase by more than a preset ratio; radio-side resources are in a state of high occupancy for a long period of time; and the cumulative number of slice service default events exceeds the target threshold.

[0118] In step 630, in response to determining that a re-orchestration has been triggered, the O-RAN slice orchestration method is re-executed.

[0119] Specifically, after a re-orchestration is triggered, the orchestration agent can re-execute the above... Figure 2The method shown can begin at either step 210 or step 230 to re-solve the current business intent, network status, cloud platform status, and service assurance risks, and generate a new deterministic orchestration result. Through this mechanism, the present invention achieves closed-loop control of the entire process from intent input, constraint generation, orchestration execution to feedback updates.

[0120] Therefore, it can be seen that the above-mentioned O-RAN slicing orchestration method can take the orchestration agent deployed on the Non-RT RIC as the core, and integrate the service intent parsing driven by the large language model, rule constraint generation, effective bandwidth modeling of burst traffic, dynamic computing jitter modeling of the cloud platform, probabilistic latency guarantee solution based on random network calculus, joint deployment and computing power allocation optimization, and O-RAN standard interface execution and feedback update into a unified closed-loop orchestration mechanism. This mechanism can not only support operators to input slice service requirements through natural language or structured templates, but also convert the requirements into verifiable mathematical constraints that meet latency and default probability requirements, and further map them into deterministic orchestration results for O-DU, O-CU deployment and CPU quota configuration. By introducing basic computing jitter measurement and load-related dynamic jitter modeling, this application can more accurately characterize the computing power fluctuations in the multi-tenant cloud environment and avoid SLA distortion caused by existing fixed service rate assumptions. By deriving the minimum continuous CPU resource lower limit based on effective bandwidth and probabilistic latency boundaries, this application can reduce the computing power waste caused by peak reservation while ensuring slice latency and reliability requirements. By mapping the solution results to the O2, O1, A1, and E2 interfaces for execution, this application enables end-to-end closed-loop control of cloud-side instance orchestration, network element configuration, and radio-side scheduling. Furthermore, it can continuously update model parameters and re-orchestration strategies based on real-time operational feedback. Therefore, the slicing orchestration scheme presented in this application possesses strong interpretability, verifiability, engineering feasibility, and system scalability, making it particularly suitable for O-RAN network slicing scenarios in multi-layer cloud environments with stringent latency and reliability requirements.

[0121] Corresponding to the above-described O-RAN slice arrangement method, embodiments of this disclosure also provide an O-RAN slice arrangement apparatus. Figure 7 An example of the internal structure of an O-RAN slicing orchestration device is shown. Figure 7 As shown, the above-mentioned O-RAN slicing and orchestration apparatus may include:

[0122] The request receiving module 710 is used to receive slice service requests; Parsing module 720 is used to generate a structured constraint object based on the slice service request; The status acquisition module 730 is used to acquire the system status set; Resource configuration module 740 is used to determine the minimum contiguous CPU resource allocation conditions required to meet the slice service level agreement requirements; and The orchestration module 750 is used to determine the slice orchestration result based on the structured constraint object, the system state set, and the minimum continuous CPU resource allocation condition; wherein the slice orchestration result includes: user association relationship, open distributed unit deployment location, open centralized unit deployment location, and CPU resource allocation scheme.

[0123] It should be noted that each module of the aforementioned O-RAN slicing orchestration device performs the above-mentioned... Figure 2-5 The specific implementation of each module in the corresponding O-RAN slice orchestration method can be found in the detailed description of each step in the O-RAN slice orchestration method in the aforementioned example. It can also achieve the technical effects that the O-RAN slice orchestration method in the aforementioned example can achieve, and will not be repeated here.

[0124] Based on the same inventive concept, corresponding to any of the above methods, an electronic device is also provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the above O-RAN slice arrangement methods.

[0125] Figure 8 A schematic diagram illustrating a more specific hardware structure of an electronic device is shown. This device may include: a processor 2010, a memory 2020, an input / output interface 2030, a communication interface 2040, and a communication interface 2050. The processor 2010, memory 2020, input / output interface 2030, and communication interface 2040 are internally connected to each other via a bus 2050.

[0126] The processor 2010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and achieve the above-mentioned technical solutions.

[0127] The memory 2020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 2020 can store the operating system and other applications. When the above technical solutions are implemented through software or firmware, the relevant program code is stored in the memory 2020 and is called and executed by the processor 2010.

[0128] The input / output interface 2030 is used to connect input / output devices to enable information input and output. These input / output devices can be configured within the device or externally connected to provide corresponding functions. Input devices may include microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0129] The communication interface 2040 is used to connect a communication component (not shown in the figure) to enable communication between this device and other devices. The communication component can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0130] Bus 2050 includes a pathway for transmitting information between various components of the device (e.g., processor 2010, memory 2020, input / output interface 2030, and communication interface 2040).

[0131] It should be noted that although the above-described device only shows the processor 2010, memory 2020, input / output interface 2030, communication interface 2040, and bus 2050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the above scheme, and does not necessarily include all the components shown in the figures.

[0132] The aforementioned electronic device is used to implement the corresponding O-RAN slice orchestration method in any of the preceding examples, and has the beneficial effects of the corresponding method, which will not be elaborated here.

[0133] Based on the same inventive concept, corresponding to any of the above example methods, in some cases a non-transitory computer-readable storage medium is also provided, which stores computer instructions for causing a computer to execute the O-RAN slicing arrangement method as described above.

[0134] The aforementioned computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program components, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0135] The computer instructions stored in the aforementioned storage medium are used to cause the computer to execute the O-RAN slicing arrangement method as described in any of the above examples, and have the corresponding beneficial effects of the method, which will not be elaborated further here.

[0136] In some cases, a computer program product is also provided, including computer program instructions, which, when run on a computer, cause the computer to execute the above-described O-RAN slicing arrangement method and have the corresponding beneficial effects, which will not be elaborated here.

[0137] Those skilled in the art should understand that the discussion of any of the above examples is merely illustrative and is not intended to imply that the scope of the claimed protection is limited to these examples; under the disclosed technical concept, the technical features of the above examples or different examples can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the examples, which are not provided in detail for the sake of brevity.

[0138] Additionally, to simplify the description and discussion, and to avoid obscuring the above examples, the provided figures may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, boards may be shown in block diagram form to avoid obscuring the examples, and this also takes into account the fact that the details of implementation of these block diagram boards are highly dependent on the platform on which these examples will be implemented (i.e., these details should be entirely within the understanding of those skilled in the art). In the context of the specific details (e.g., circuits) set forth to describe these examples, it will be apparent to those skilled in the art that the above examples can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0139] Although various examples have been described, many substitutions, modifications, and variations of these examples will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) can use the examples discussed.

[0140] These examples are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of these examples shall be included within the scope of protection.

Claims

1. An open radio access network slicing orchestration method, comprising: Receive slice service requests; A structured constraint object is generated based on the slice service request; Obtain the system state set; Determine the minimum contiguous CPU resource allocation conditions that meet the requirements of the slice service level agreement; as well as Based on the structured constraint object, the system state set, and the minimum continuous CPU resource allocation condition, the slice orchestration result is determined; wherein, the slice orchestration result includes: user association, open distributed unit deployment location, open centralized unit deployment location, and CPU resource allocation scheme.

2. The open radio access network slicing orchestration method according to claim 1, wherein, The generation of structured constraint objects based on the slice service request includes: In response to determining that the slice service request is a slice service request in natural language form, a large language model is used to perform semantic understanding and key information extraction on the slice service request, generating a structured constraint object; wherein, the structured constraint object includes: slice type, service area, end-to-end latency budget, latency default probability threshold, priority, deployment preference, and migration strategy, or any combination thereof; or, In response to determining that the slice service request is a slice service request in the form of a structured template, the slice service request is parsed based on the predefined structured template to generate the structured constraint object.

3. The open radio access network slicing and orchestration method according to claim 1, further comprising: The structured constraint object is validated according to predefined validation rules to obtain a structured constraint object that conforms to the validation rules; wherein, the validation rules include one or any combination of parameter range checking, field completion and illegal request filtering.

4. The open radio access network slicing and orchestration method according to claim 1, wherein, The system status set includes: Open Radio Access Network (O-RAN) status information, cloud platform telemetry information, access network intelligent controller (RIC) information, and auxiliary information, or any combination thereof. The system state set to be acquired includes: The O-RAN status information is obtained through the operation and maintenance management interface O1; wherein, the O-RAN status information includes: O-RAN network element management status, network topology and radio resource occupancy status, radio coverage status, link reachability, and alarm information, or any combination thereof; or, The cloud platform telemetry information is obtained through the business orchestration automation interface O2; wherein, the cloud platform telemetry information may include: CPU available capacity, container instance running status, resource usage, node health, instance location, and processing latency sampling information, or any combination thereof; or, The RIC information is obtained through a near real-time wireless access network intelligent controller or monitoring module; wherein the RIC information includes: wireless load summary, user access status, handover status, and one or any combination thereof of Open Fronthaul Intelligent Interface (E2) execution feedback; or, Collect the auxiliary information; wherein the auxiliary information includes: the server's maximum allocatable CPU resources, the wireless resource block capacity, the connection relationship between the server and the open radio frequency unit, the user's current access location, and the existing network element deployment status, or any combination thereof.

5. The open radio access network slicing and orchestration method according to claim 1, wherein, The minimum contiguous CPU resource allocation conditions that satisfy the requirements of the slice service level agreement include: Determine the server Slice-oriented Example of the function Aggregated bandwidth; Determine the server Slice-oriented Dynamic calculation of jitter parameters; and The server is determined based on the aggregated bandwidth and the dynamically calculated jitter parameters. Slice-oriented Example of the function The minimum amount of contiguous CPU resources allocated; and The minimum amount of consecutive CPU resources is used as the minimum consecutive CPU resource allocation condition.

6. The open radio access network slicing and orchestration method according to claim 5, wherein, The determining server Slice-oriented Example of the function The aggregated bandwidth includes: Based on expression Determine user effective bandwidth ;in, Represents the state transition matrix; Represents the arrival rate matrix; This represents the stringency parameter of the constraint. Indicates the spectral radius; and For deployment on servers upper slice Functional Examples The server is obtained by summing the effective bandwidth of each user in the user set. Slice-oriented Example of the function aggregated bandwidth .

7. The open radio access network slicing and orchestration method according to claim 5, wherein, The server is determined Slice-oriented The dynamic calculation of jitter parameters includes: The processing latency of the protocol stack tasks is sampled within a preset statistical window to obtain the server's latency. Processing delay sample sequence within the statistical window ; Based on expression Determine the server Basic calculation of jitter parameters ;in, Represents the standard deviation operation; and Based on the server Basic calculation of jitter parameters and expressions Determine the server Slice-oriented Dynamic calculation of jitter parameters ;in, Indicates the server The slice currently being carried The number of users; Indicates the server The load sensitivity coefficient.

8. The open radio access network slice orchestration method according to claim 5, wherein, Based on the server Slice-oriented Example of the function The aggregated bandwidth and the server Slice-oriented The dynamic calculation of jitter parameters determines the server Slice-oriented Example of the function The minimum amount of contiguous CPU resources allocated includes: The forward and midhaul delays are determined based on the system state set. Extract the end-to-end delay budget from the structured constraint object; The total computational delay budget is determined based on the end-to-end delay budget, the fronthaul delay, and the midhaul delay. ; Based on the total computational delay budget Determine functional instances Local computational delay budget ;as well as Based on expression Determine the server Slice-oriented Example of the function Minimum amount of contiguous CPU resources allocated ;in, This indicates the processing efficiency of the O-RAN protocol stack; This indicates the CPU's fixed clock frequency; Representing the server Slice-oriented Example of the function Aggregated bandwidth; Representing the server Slice-oriented Dynamic calculation of jitter parameters; This represents the stringency parameter of the constraint. This represents the default probability threshold in the structured constraint object.

9. The open radio access network slicing orchestration method according to claim 8, wherein, The determination of the slice orchestration result based on the structured constraint object, the system state set, and the minimum contiguous CPU resource allocation condition includes: Generate a set of candidate paths for each user based on the network topology relationships in the system state set; Based on the wireless coverage relationship, link reachability relationship, server resource status and end-to-end latency budget in the system state set, the candidate path set is pre-pruned to obtain the feasible candidate path set for each user. An initial orchestration population is generated based on the set of feasible candidate paths for each user. Decode the individuals in the initial orchestration population to obtain the user-related variables, open distributed unit deployment variables, and open centralized unit deployment variables corresponding to each individual; Based on the user associations, the deployment locations of the open distributed units, and the deployment locations of the open centralized units, determine the set of slice users carried on each server; Based on the slice user set, the server allocates central processing unit resources to functional instances on the slice; Constraint verification is performed on each individual in the initial orchestration population based on the central processing unit resource allocation results. Determine the fitness value of each individual in the initial population arrangement; Based on the fitness values ​​of each individual in the initial staging population, the initial staging population is updated. Return to the step of decoding individuals in the initial orchestration population until the iteration termination condition is met; From the updated choreography population or the historical best individuals, select the individual with the best fitness value that meets the constraints as the target choreography individual; and Based on the stated target, the individual outputs a deterministic slice arrangement result.

10. The open radio access network slicing orchestration method according to claim 1, further comprising: Based on the deployment locations of the open distributed units, the deployment locations of the open centralized units, and the CPU resource allocation scheme, the requests are converted into O2 orchestration requests. The O2 orchestration request is sent to the open cloud infrastructure platform via the business orchestration automation interface O2. Instantiation of the target open distributed unit and the target open centralized unit is performed based on the deployment locations of the open distributed unit and the open centralized unit. Set the CPU quotas for the target open distributed unit and the target open centralized unit; Initialization configurations are issued to the target open distributed unit and the target open centralized unit through the operation and maintenance management interface O1; wherein, the initialization configurations include: network element identifier, network layer address, adjacent network element information, bearer connection parameters, slice instance identifier, service link configuration parameters or any combination thereof; A1 policy information is generated based on user association relationships; The A1 policy information is sent to the near real-time wireless access network intelligent controller via the wireless intelligent management interface A1. Generate scheduling guidance information based on A1 policy information; and The scheduling guidance information is sent to the target open distributed unit and related E2 nodes through the open fronthaul intelligent interface E2.

11. The open radio access network slicing orchestration method according to claim 1, further comprising: Obtain feedback information on slice arrangement; Whether to trigger re-arrangement is determined based on the slice arrangement feedback information and the preset re-arrangement conditions; as well as In response to determining that the re-orchestration is triggered, the Open Radio Access Network (RAN) slice orchestration method is re-executed.

12. The open radio access network slicing orchestration method according to claim 11, wherein, The acquisition of slice arrangement feedback information includes: O-RAN feedback information is obtained through the operation and maintenance management interface O1; wherein, the O-RAN feedback information includes one or any combination of information such as network element alarms, connection status, and configuration change results; or, Cloud platform feedback information is obtained through the business orchestration automation interface O2; wherein, the cloud platform feedback information includes: cloud platform instance running status, CPU utilization, resource contention, container throttling events, and processing latency statistics, or any combination thereof; or, The feedback information from the wireless access network intelligent controller is obtained through the near real-time wireless access network intelligent controller or its supporting monitoring module; wherein, the feedback information from the wireless access network intelligent controller may include one or any combination of information such as wireless resource utilization, changes in user access, slice service quality indicators, and execution results of the open fronthaul intelligent interface E2.

13. The open radio access network slicing orchestration method according to claim 11, wherein, The reordering conditions include: The actual processing latency of the sliced ​​business continuously approaches or exceeds the local budget threshold; or, The number of CPU throttling events on the server exceeds a preset threshold; or, The increase in the basic or dynamically calculated jitter parameter exceeds a preset proportional threshold; or, Wireless resources are consistently under high occupancy; or, The cumulative number of default events in the slicing business has exceeded the pre-set target threshold.

14. An electronic device comprising: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the open radio access network slicing orchestration method as described in any one of claims 1-13.

15. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the open radio access network slicing and orchestration method according to any one of claims 1-13.

16. A computer program product comprising computer program instructions that, when executed on a computer, cause the computer to perform the open radio access network slicing orchestration method as described in any one of claims 1-13.

Images