CPE service distribution management method and system based on 5G network slicing

Abstract

The invention discloses a CPE service distribution management method and system based on 5G network slices, and relates to the technical field of flow management. The method comprises the following steps: collecting application layer identification data and network layer feature data through CPE equipment, and carrying out service type identification on the application layer identification data and the network layer feature data by adopting a service classifier to obtain CPE service type information; dynamically allocating the 5G network slice resources, and determining network slice resource allocation parameters; traffic monitoring prediction is carried out on the CPE service type information by adopting a service traffic prediction channel, and CPE service traffic prediction parameters are output; and performing shunting optimization on the network slice resource allocation parameters, configuring service network slice resource parameters, and performing service shunting closed-loop management. The technical problem that in the prior art, 5G network slice resources are difficult to dynamically allocate according to different services, and consequently the resource utilization rate is low is solved, and the technical effect of improving the resource utilization rate is achieved.

Description

Technical Field

[0001] This invention relates to the field of traffic management technology, specifically to a CPE service offloading management method and system based on 5G network slicing. Background Technology

[0002] With the rapid development of 5G communication technology, network slicing, as one of the important capabilities of 5G networks, is widely used to support differentiated service needs. By constructing multiple logically isolated network slices on the same physical network infrastructure, it meets the service quality requirements of different services in terms of bandwidth, latency, and reliability. In actual deployments, users typically access the 5G network through Customer Premises Equipment (CPE), and traffic from different applications converges at the CPE side and is transmitted via network slices. However, in existing technologies, the management of service traffic on the CPE side often relies on static configuration or simple rules based on ports and protocols, making it difficult to accurately identify complex and diverse service types and to promptly detect dynamic changes in service traffic. Simultaneously, the allocation of 5G network slice resources often adopts pre-configuration or fixed mapping methods, lacking a coordination mechanism with specific service types and their real-time traffic characteristics. This results in some network slice resources being idle while others are congested under fluctuating service loads or multi-service concurrent scenarios, leading to low overall resource utilization. Summary of the Invention

[0003] This application provides a CPE service offloading management method and system based on 5G network slicing, which solves the technical problem in the prior art that it is difficult to dynamically allocate 5G network slice resources according to different services, resulting in low resource utilization.

[0004] The first aspect of this application provides a CPE service offloading management method based on 5G network slicing, the method comprising: Application layer identification data and network layer feature data are collected through CPE devices. A service classifier is used to identify the service type of the application layer identification data and network layer feature data to obtain CPE service type information. 5G network slice resources are acquired, and the 5G network slice resources are dynamically allocated based on the CPE service type information to determine the network slice resource allocation parameters. A service traffic prediction channel is established, and the traffic prediction channel is used to monitor and predict the traffic of the CPE service type information, outputting CPE service traffic prediction parameters. A service traffic offloading strategy engine optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters, configures the service network slice resource parameters, and performs closed-loop management of service traffic offloading through the service network slice resource parameters.

[0005] A second aspect of this application provides a CPE service offloading management system based on 5G network slicing, the system comprising: Data Acquisition Module: Collects application layer identification data and network layer feature data through CPE devices, and uses a service classifier to identify the service type of the application layer identification data and network layer feature data to obtain CPE service type information; Resource Allocation Module: Acquires 5G network slice resources, dynamically allocates the 5G network slice resources based on the CPE service type information, and determines the network slice resource allocation parameters; Traffic Prediction Module: Establishes a service traffic prediction channel, uses the service traffic prediction channel to perform traffic monitoring and prediction on the CPE service type information, and outputs CPE service traffic prediction parameters; Traffic Splitting Management Module: Optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters through a service traffic splitting strategy engine, configures service network slice resource parameters, and performs closed-loop management of service traffic splitting through the service network slice resource parameters.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, application layer identification data and network layer feature data are collected through CPE devices. A service classifier is then used to identify the service types of the application layer identification data and network layer feature data to obtain CPE service type information. Next, 5G network slice resources are acquired, and these resources are dynamically allocated based on the CPE service type information to determine the network slice resource allocation parameters. Then, a service traffic prediction channel is established to monitor and predict traffic based on the CPE service type information, outputting CPE service traffic prediction parameters. Finally, a service offloading strategy engine optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters, configures the service network slice resource parameters, and performs closed-loop management of service offloading using these parameters. This solves the technical problem in existing technologies where it is difficult to dynamically allocate 5G network slice resources according to different services, leading to low resource utilization, and achieves the technical effect of improving resource utilization. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram of the CPE service offloading management method based on 5G network slicing provided in this application embodiment; Figure 2 A schematic diagram of the CPE service offloading management system based on 5G network slicing provided in this application embodiment.

[0009] Figure labeling: Data acquisition module 11, resource allocation module 12, traffic prediction module 13, traffic diversion management module 14. Detailed Implementation

[0010] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0011] Example 1, as Figure 1 As shown, this application provides a CPE service offloading management method based on 5G network slicing, wherein the method includes: Application layer identification data and network layer feature data are collected by CPE devices, and a service classifier is used to identify the service type of the application layer identification data and network layer feature data to obtain CPE service type information.

[0012] In this embodiment, a CPE device deployed on the user side collects the data stream carrying the service in real time. The application layer identification data includes at least one of the following: service port number, application protocol type, application identifier field, session identifier, and service request feature code. The network layer feature data includes at least one of the following: packet 5-tuple information, packet length distribution, average throughput rate, latency, packet loss rate, and connection duration. The CPE device preprocesses the collected application layer identification data and network layer feature data. This preprocessing includes invalid data filtering, duplicate data removal, and unified timestamp processing to ensure consistency between different data sources within the same time window. Subsequently, the CPE device performs feature extraction on the preprocessed data, mapping the application layer identification data and network layer feature data into numerical features, categorical features, and temporal features, respectively. The different types of features are then standardized to construct a key feature vector for the service application network. This key feature vector is input into a service classifier pre-trained based on historical CPE service data. The service classifier performs inference calculations on the key feature vector and outputs the corresponding service type label, thereby obtaining the CPE service type information. The service type information is used to characterize the service category to which the current CPE service belongs, and serves as the basic input parameter for subsequent 5G network slice resource allocation and service offloading management.

[0013] Furthermore, obtaining CPE service type information includes: The application layer identification data and network layer feature data are subjected to invalid data filtering, duplicate data removal, and timestamp standardization to obtain standard application layer identification data and standard network layer feature data. Numerical, categorical, and temporal features are extracted from the standard application layer identification data and standard network layer feature data to obtain a key feature set of the business application network. A business classifier is constructed based on the historical CPE business dataset for classification training. The business classifier is used to identify the business type of the key feature set of the business application network to obtain CPE business type information.

[0014] First, the application layer identification data and network layer feature data collected by the CPE device are preprocessed. This preprocessing includes filtering outliers, missing values, and data records without business significance; removing duplicate data collected repeatedly within the same time window; and standardizing and calibrating timestamps from different data sources to eliminate the impact of clock skew on subsequent analysis, thus obtaining standard application layer identification data and standard network layer feature data. Second, feature extraction is performed on the standard application layer identification data and standard network layer feature data. This feature extraction includes converting continuous indicators into numerical features, converting discrete information such as protocol type and port category into categorical features, and modeling time-series features such as changes in business traffic and connection persistence according to a preset time window, thereby forming a set of key network features for business application behavior. Then, a business classification model is trained based on a pre-constructed historical CPE business dataset. This historical CPE business dataset contains historical business feature samples with completed business type labeling. Supervised training is performed on these historical business feature samples to construct a business classifier capable of distinguishing and recognizing different business types. Finally, the key feature set of the business application network is input into the business classifier for inference calculation. The business classifier outputs the corresponding business type identification result as the business type label of the current CPE business, thereby obtaining the CPE business type information.

[0015] Furthermore, constructing a business classifier includes: The CPE business historical dataset is characterized by feature identification and business type labeling to obtain a CPE business type sample set. A decision tree is used to calculate the information gain and recursively split the features of each sample in the CPE business type sample set to construct an initial classification model. Nodes in the initial classification model are backtracked upwards, and error evaluation is performed to prune the model, resulting in a post-processing classification model. Performance verification and optimization are performed based on the post-processing classification model to minimize the model loss function, thus obtaining a business classifier.

[0016] First, the historical CPE service dataset is processed to construct samples. This dataset includes application layer identifier data, network layer feature data, and corresponding service type labels for historical services. Each service sample in the historical CPE service dataset is characterized by features, and service type annotations are completed by combining service source, service protocol type, or manual verification results, thus forming a CPE service type sample set containing feature vectors and corresponding service type labels. Second, an initial decision tree classification model is constructed based on the CPE service type sample set. Specifically, the decision tree algorithm is used to calculate the information gain of each feature dimension in the CPE service type sample set. The feature with the largest information gain is used as the splitting feature of the current node, and the sample set is recursively split according to the value range or category of this feature until a preset stopping condition is met. The stopping condition includes at least one of a sample purity threshold, maximum tree depth, or minimum number of samples, thereby constructing the initial classification model. Subsequently, a post-processing pruning operation is performed on the initial classification model. This pruning operation includes backtracking analysis of each non-leaf node in the decision tree, evaluating the classification error before and after node splitting based on validation samples, and pruning the corresponding branch when node splitting fails to significantly reduce the overall classification error, thereby reducing model complexity and obtaining the post-processed classification model. Finally, performance verification and parameter tuning are performed based on the post-processed classification model. By inputting validation samples into the post-processed classification model, the classification accuracy, misclassification rate, or loss function value is calculated, and the model parameters are adjusted according to the performance evaluation results to minimize the model loss function, ultimately obtaining a business classifier for CPE business type identification.

[0017] Acquire 5G network slice resources, dynamically allocate the 5G network slice resources based on the CPE service type information, and determine the network slice resource allocation parameters.

[0018] In this embodiment, a communication connection is established with the network management function unit on the 5G core network side to obtain currently available 5G network slice resource information. This 5G network slice resource information includes the slice identifier, available bandwidth resources, latency capability, reliability level, and current load status parameters for each network slice. Based on the obtained 5G network slice resource information, combined with the CPE service type information obtained in the preceding steps, the compatibility relationship between different service types and network slice resources is analyzed. According to the CPE service type information, a corresponding network slice candidate set is selected from the preset service type-network slice capability mapping relationship. Based on the differences in bandwidth, latency, and stability requirements of services, the slice resources in the network slice candidate set are dynamically allocated and adjusted. The dynamic allocation process includes configuring slice bandwidth quotas, scheduling priorities, and resource occupancy ratios to meet the carrying requirements of the current service type. Network slice resource allocation parameters are determined based on the dynamic allocation results. These parameters describe the resource occupancy method of each CPE service in the corresponding network slice and serve as the basic control parameters for subsequent service traffic prediction, traffic offloading optimization, and closed-loop management.

[0019] Furthermore, determining network slice resource allocation parameters includes: According to the CPE service scenario requirements, the 5G network slice resources are analyzed for slice configuration to generate a service scenario type-slice resource configuration parameter template library. Based on the CPE service type information, the template library is matched and selected to obtain an initial slice resource parameter template. The slice resource status parameters are monitored in real time by the SDN controller. If the slice resource status parameters are lower than a preset resource threshold, a dynamic adjustment mechanism is triggered. Based on the dynamic adjustment mechanism, the initial slice resource parameter template is dynamically allocated to determine the network slice resource allocation parameters.

[0020] Based on the application scenario requirements of different services on the CPE side, the available 5G network slice resources are analyzed for slice configuration. CPE service scenario requirements include bandwidth, latency, reliability, and concurrent connection capabilities. Based on these requirements, different service scenario types are associated with corresponding network slice capability parameters. Services are categorized according to their scenarios, which include at least low-latency interactive services, high-bandwidth transmission services, stable connection services, and ordinary data transmission services. For each service scenario, corresponding slice resource parameters are pre-configured, including at least bandwidth limits, bandwidth guarantees, latency levels, packet loss rate thresholds, and scheduling priorities. This establishes a mapping relationship between service scenario types and slice resource configuration parameters, generating a service scenario type-slice resource configuration parameter template library.

[0021] Based on CPE service type information, a matching selection is made from the service scenario type-slice resource configuration parameter template library. A corresponding initial slice resource parameter template is selected for each type of CPE service, ensuring the selected template meets the basic bearer requirements of that service type, thus obtaining the initial slice resource parameter template. Subsequently, the SDN controller monitors the resource usage status of each network slice in real time. The slice resource status parameters include remaining bandwidth, current load rate, latency fluctuation, and packet loss. When any slice resource status parameter is detected to be lower than a preset resource threshold, it is determined that the current slice resource configuration cannot meet the service bearer requirements, triggering a dynamic adjustment mechanism. This dynamic adjustment mechanism adjusts the resource configuration parameters in the initial slice resource parameter template without affecting the isolation of other slices, including increasing or decreasing slice bandwidth quotas, adjusting scheduling priorities, or reallocating services to backup slices. After the dynamic adjustment mechanism is triggered, the initial slice resource parameter template is dynamically allocated and adjusted. Based on real-time slice resource status and changes in service bearer requirements, the slice resource parameters are updated and configured to determine the network slice resource allocation parameters. These network slice resource allocation parameters are used for subsequent service offloading optimization and closed-loop management.

[0022] Furthermore, based on the aforementioned dynamic adjustment mechanism, the initial slice resource parameter template is dynamically allocated to determine the network slice resource allocation parameters, including: Based on the dynamic adjustment mechanism, a multi-dimensional service priority evaluation system is obtained; the priority of each service type in the CPE service type information is evaluated according to the multi-dimensional service priority evaluation system to generate a service type priority evaluation matrix; the initial slice resource parameter template is dynamically adjusted and allocated based on the service type priority evaluation matrix to determine the network slice resource allocation parameters.

[0023] After the dynamic adjustment mechanism is triggered, a pre-built multi-dimensional service priority evaluation system is invoked. The multi-dimensional service priority evaluation system includes at least the service service level requirement dimension, the real-time traffic load dimension, the service latency sensitivity dimension, and the service historical stability dimension. Each dimension is set with a configurable weight coefficient to reflect the relative demand intensity of different services for slice resources under the current network operating state.

[0024] Based on a multi-dimensional business priority evaluation system, priority evaluation is performed on each business type in the CPE business type information. Specifically, for each business type, its corresponding business feature parameters are extracted, and a normalized priority score is calculated according to each evaluation dimension. The priority scores of each dimension are weighted and summed according to their corresponding weight coefficients to obtain a comprehensive priority score for each business type. A business type priority evaluation matrix is ​​constructed based on the comprehensive priority score. The business type priority evaluation matrix is ​​used to represent the resource scheduling priority order and relative weight relationship of different business types at the current time.

[0025] The initial slice resource parameter template is dynamically adjusted and allocated based on the service type priority evaluation matrix. Specifically, this includes: sorting each service type according to the service type priority evaluation matrix, and determining the proportion adjustment factor of each service type in slice resource allocation based on the sorting results; and dynamically adjusting resource parameters such as bandwidth quota, scheduling priority, or connection limit in the initial slice resource parameter template in combination with the proportion adjustment factor, thereby generating network slice resource allocation parameters that match the current service priority distribution, thus realizing differentiated and dynamic allocation of network slice resources.

[0026] Furthermore, based on the service type priority evaluation matrix, the initial slice resource parameter template is dynamically adjusted and allocated to determine the network slice resource allocation parameters, including: Based on the service type priority evaluation matrix, service types are sorted to obtain a service type priority sequence; based on the service type priority sequence, a service type traffic allocation decision ratio is determined; based on the service type traffic allocation decision ratio, the initial slice resource parameter template is dynamically adjusted and allocated to determine the network slice resource allocation parameters.

[0027] Preferably, the service types are sorted according to the comprehensive priority score corresponding to each service type in the service type priority evaluation matrix, and a service type priority sequence is generated according to the priority score from high to low; the service type priority sequence is used to characterize the resource scheduling order of each service type under the current network state.

[0028] After obtaining the service type priority sequence, the service type traffic allocation decision ratio is determined based on the service type priority sequence. Specifically, according to the priority score difference between adjacent service types in the service type priority sequence, combined with a preset ratio mapping rule or normalization rule, the comprehensive priority score of each service type is converted into a corresponding traffic allocation weight coefficient. The traffic allocation weight coefficient is normalized so that the sum of the traffic allocation decision ratios corresponding to each service type is a preset total ratio, thereby obtaining a set of service type traffic allocation decision ratios. The service type traffic allocation decision ratio is used to characterize the resource occupancy ratio of each service type in network slice resource allocation.

[0029] The initial slice resource parameter template is dynamically adjusted and allocated based on the service type diversion decision ratio. Specifically, this includes: adjusting the bandwidth quota, scheduling weight, or connection limit in the initial slice resource parameter template proportionally according to the service type diversion decision ratio; when there are spare slice resources, migrating some service traffic to the spare slice according to the service type diversion decision ratio; generating the final network slice resource allocation parameters through the above adjustments, and using the network slice resource allocation parameters for subsequent service diversion execution.

[0030] Establish a business traffic prediction channel, use the business traffic prediction channel to perform traffic monitoring and prediction on the CPE business type information, and output CPE business traffic prediction parameters.

[0031] In this embodiment of the application, a business traffic prediction channel is established to continuously monitor and predict traffic changes of different business types on the CPE side.

[0032] During business operation, the real-time business traffic data corresponding to the current moment is input into the business traffic prediction channel. According to the preset prediction period, the business traffic of each business type in the future time period is predicted, and the corresponding CPE business traffic prediction parameters are output. The CPE business traffic prediction parameters include at least the predicted traffic value in the future time window, the traffic change trend index, and the peak traffic prediction result. The prediction parameters serve as the input basis for subsequent dynamic allocation of network slice resources and optimization of business diversion strategies.

[0033] Furthermore, establishing a business traffic prediction channel includes: Collect historical business traffic datasets, clean and timestamp the data to obtain a business traffic time series dataset; obtain a set of factors influencing business traffic, and label the business traffic time series dataset according to the set of factors influencing business traffic and a preset time window to obtain a business traffic time series sample set; use a time series network to train traffic prediction on the business traffic time series sample set to establish a business traffic prediction channel.

[0034] First, historical service traffic data from the CPE side is continuously collected according to service type to construct a historical service traffic dataset. This historical data includes at least statistical indicators such as uplink traffic, downlink traffic, total traffic, number of sessions, and traffic fluctuation amplitude collected over a continuous time period. The historical service traffic dataset undergoes data cleaning, including outlier removal, missing data completion, and duplicate data removal. A timestamp is added to the cleaned data to form a time-series service traffic dataset arranged chronologically. After obtaining the time-series dataset, a set of service traffic influencing factors related to service traffic changes is acquired. This set includes at least service type, historical traffic change trends, time period characteristics, and network load status parameters. Based on this set of influencing factors, the time-series dataset is segmented according to a preset time window. Influencing factors are labeled for the service traffic data within each time window to generate a corresponding time-series sample set. Each time-series sample contains an input feature vector and a corresponding target traffic value. Subsequently, based on the time series sample set of business traffic, the preset time series network is trained for traffic prediction. The network parameters are iteratively updated to minimize the prediction error until the preset convergence condition is met, thus completing the construction of the business traffic prediction model. The trained time series network serves as a business traffic prediction channel, used to predict the changes in business traffic of different business types within future time windows, and to provide a forward-looking traffic prediction basis for subsequent dynamic allocation of network slice resources and optimization of business diversion strategies.

[0035] The service traffic allocation engine optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters, configures the service network slice resource parameters, and performs closed-loop management of service traffic allocation through the service network slice resource parameters.

[0036] In this embodiment, a service offloading strategy engine optimizes network slice resources and implements closed-loop management. First, the service offloading strategy engine receives CPE service traffic prediction parameters and network slice resource allocation parameters. Based on the predicted future traffic demands of each service type, it evaluates the existing network slice resource allocation status and identifies potential resource shortages or idle resources. Then, according to a preset slice resource offloading strategy, the service offloading strategy engine performs offloading optimization processing on the network slice resource allocation parameters. This optimization process includes adjusting the carrying ratio of different service types across multiple network slices and reconfiguring slice bandwidth quotas and scheduling priorities, thereby generating service network slice resource parameters. These parameters indicate the carrying method and resource occupancy relationship of each service flow in different network slices.

[0037] During service operation, service traffic splitting control is executed based on service network slice resource parameters, and the splitting effect is continuously monitored to obtain real-time slice resource usage status and service carrying feedback information. The service network slice resource parameters are dynamically updated based on the service carrying feedback information, and service traffic splitting is executed again using the updated parameters, thus forming a closed-loop management process for service traffic splitting. This allows network slice resource allocation to continuously adapt to changes in service traffic, improving overall resource utilization efficiency and service carrying stability.

[0038] Furthermore, configuring service network slice resource parameters includes: According to the service traffic offloading strategy engine, the slice resource offloading strategy library is invoked. The slice resource offloading strategy library includes priority offloading rules, peak period offloading rules, and backup slice offloading rules. Based on the CPE service traffic prediction parameters, resource gap calculation is performed to obtain CPE service traffic resource gap parameters. According to the slice resource offloading strategy library, the network slice resource allocation parameters are optimized based on the CPE service traffic resource gap parameters to configure the service network slice resource parameters.

[0039] After the business traffic splitting strategy engine starts, it calls the pre-built slice resource splitting strategy library. The slice resource splitting strategy library is used to define slice resource splitting rules under different business scenarios. The slice resource splitting rules include at least priority splitting rules based on business type priority, peak period splitting rules for business traffic surge scenarios, and backup slice splitting rules activated when the primary slice resources are insufficient.

[0040] After obtaining the CPE service traffic prediction parameters, the service traffic distribution strategy engine compares the predicted traffic value of each service type within the prediction time window with the available resources of the corresponding slice in the current network slice resource allocation parameters to calculate the difference between the resource demand and available resources for each service type. This difference is used as the CPE service traffic resource gap parameter. The resource gap parameter is used to characterize the degree to which the current slice resource configuration cannot meet the service traffic demand within the prediction time window.

[0041] According to the traffic allocation rules in the slice resource allocation strategy library, the network slice resource allocation parameters are optimized based on the CPE service traffic resource gap parameter. Specifically, this includes: when the resource gap parameter corresponding to a service type is lower than a preset threshold, maintaining the original slice resource allocation; when the resource gap parameter exceeds the preset threshold, prioritizing slice resources for high-priority services according to priority traffic allocation rules; during peak hours, limiting or cross-slice traffic for some low-priority services according to peak hour traffic allocation rules; and when the primary slice resources still cannot meet the demand, migrating some service traffic to the backup slice according to the backup slice traffic allocation rules. Through the above traffic optimization, service network slice resource parameters are generated for actual service traffic allocation execution.

[0042] Furthermore, closed-loop management of service traffic distribution through the aforementioned service network slice resource parameters includes: Real-time monitoring of service traffic distribution is performed using the service network slice resource parameters to obtain slice resource status feedback parameters; the service network slice resource parameters are optimized and updated based on the slice resource status feedback parameters, and closed-loop management of service traffic distribution is performed using the updated service network slice resource parameters.

[0043] During the service offloading execution phase, the data streams of each service type are sliced ​​and controlled based on the configured service network slice resource parameters, and the offloading process is monitored in real time. The real-time monitoring includes at least the collection of operating status parameters such as bandwidth utilization, service traffic completion rate, link latency, packet loss rate, and slice resource utilization of each network slice. The collected operating status parameters are used as slice resource status feedback parameters.

[0044] After obtaining the slice resource status feedback parameters, these parameters are analyzed and compared with preset performance thresholds and operational targets. When any slice resource status feedback parameter deviates from the preset threshold or shows a continuous deterioration trend, it is determined that the current service network slice resource parameters cannot continuously meet the service operation requirements, triggering a parameter optimization and update process. This parameter optimization and update process is used to adjust the slice bandwidth quota, scheduling priority, or traffic offloading ratio in the service network slice resource parameters based on the slice resource status feedback parameters. After completing the parameter optimization and update, the updated service network slice resource parameters are redistributed and applied to service traffic offloading control, and the updated offloading effect is continuously monitored. Through the cyclical execution of the above real-time monitoring, feedback analysis, and parameter updates, a closed-loop management mechanism for service traffic offloading based on service network slice resource parameters is formed, thereby achieving dynamic adaptive adjustment between network slice resource configuration and service traffic changes.

[0045] In summary, the embodiments of this application have at least the following technical effects: First, application layer identification data and network layer feature data are collected through CPE devices. A service classifier is then used to identify the service types of the application layer identification data and network layer feature data to obtain CPE service type information. Next, 5G network slice resources are acquired, and these resources are dynamically allocated based on the CPE service type information to determine the network slice resource allocation parameters. Then, a service traffic prediction channel is established to monitor and predict traffic based on the CPE service type information, outputting CPE service traffic prediction parameters. Finally, a service offloading strategy engine optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters, configures the service network slice resource parameters, and performs closed-loop management of service offloading using these parameters. This solves the technical problem in existing technologies where it is difficult to dynamically allocate 5G network slice resources according to different services, leading to low resource utilization, and achieves the technical effect of improving resource utilization.

[0046] Example 2 is based on the same inventive concept as the CPE service offloading management method based on 5G network slicing in the previous examples, such as... Figure 2 As shown, this application provides a CPE service offloading management system based on 5G network slicing, wherein the system includes: Data Acquisition Module 11: Collects application layer identification data and network layer feature data through CPE devices, and uses a service classifier to identify the service type of the application layer identification data and network layer feature data to obtain CPE service type information; Resource Allocation Module 12: Acquires 5G network slice resources, dynamically allocates the 5G network slice resources based on the CPE service type information, and determines the network slice resource allocation parameters; Traffic Prediction Module 13: Establishes a service traffic prediction channel, uses the service traffic prediction channel to perform traffic monitoring and prediction on the CPE service type information, and outputs CPE service traffic prediction parameters; Traffic Splitting Management Module 14: Optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters through a service traffic splitting strategy engine, configures service network slice resource parameters, and performs closed-loop management of service splitting through the service network slice resource parameters.

[0047] Furthermore, the data acquisition module 11 is used to perform the following methods: The application layer identification data and network layer feature data are subjected to invalid data filtering, duplicate data removal, and timestamp standardization to obtain standard application layer identification data and standard network layer feature data. Numerical, categorical, and temporal features are extracted from the standard application layer identification data and standard network layer feature data to obtain a key feature set of the business application network. A business classifier is constructed based on the historical CPE business dataset for classification training. The business classifier is used to identify the business type of the key feature set of the business application network to obtain CPE business type information.

[0048] Furthermore, the data acquisition module 11 is used to perform the following methods: The CPE business historical dataset is characterized by feature identification and business type labeling to obtain a CPE business type sample set. A decision tree is used to calculate the information gain and recursively split the features of each sample in the CPE business type sample set to construct an initial classification model. Nodes in the initial classification model are backtracked upwards, and error evaluation is performed to prune the model, resulting in a post-processing classification model. Performance verification and optimization are performed based on the post-processing classification model to minimize the model loss function, thus obtaining a business classifier.

[0049] Furthermore, the resource allocation module 12 is used to perform the following method: According to the CPE service scenario requirements, the 5G network slice resources are analyzed for slice configuration to generate a service scenario type-slice resource configuration parameter template library. Based on the CPE service type information, the template library is matched and selected to obtain an initial slice resource parameter template. The slice resource status parameters are monitored in real time by the SDN controller. If the slice resource status parameters are lower than a preset resource threshold, a dynamic adjustment mechanism is triggered. Based on the dynamic adjustment mechanism, the initial slice resource parameter template is dynamically allocated to determine the network slice resource allocation parameters.

[0050] Furthermore, the resource allocation module 12 is used to perform the following method: Based on the dynamic adjustment mechanism, a multi-dimensional service priority evaluation system is obtained; the priority of each service type in the CPE service type information is evaluated according to the multi-dimensional service priority evaluation system to generate a service type priority evaluation matrix; the initial slice resource parameter template is dynamically adjusted and allocated based on the service type priority evaluation matrix to determine the network slice resource allocation parameters.

[0051] Furthermore, the resource allocation module 12 is used to perform the following method: Based on the service type priority evaluation matrix, service types are sorted to obtain a service type priority sequence; based on the service type priority sequence, a service type traffic allocation decision ratio is determined; based on the service type traffic allocation decision ratio, the initial slice resource parameter template is dynamically adjusted and allocated to determine the network slice resource allocation parameters.

[0052] Furthermore, the traffic prediction module 13 is used to perform the following method: Collect historical business traffic datasets, clean and timestamp the data to obtain a business traffic time series dataset; obtain a set of factors influencing business traffic, and label the business traffic time series dataset according to the set of factors influencing business traffic and a preset time window to obtain a business traffic time series sample set; use a time series network to train traffic prediction on the business traffic time series sample set to establish a business traffic prediction channel.

[0053] Furthermore, the traffic management module 14 is used to perform the following methods: According to the service traffic offloading strategy engine, the slice resource offloading strategy library is invoked. The slice resource offloading strategy library includes priority offloading rules, peak period offloading rules, and backup slice offloading rules. Based on the CPE service traffic prediction parameters, resource gap calculation is performed to obtain CPE service traffic resource gap parameters. According to the slice resource offloading strategy library, the network slice resource allocation parameters are optimized based on the CPE service traffic resource gap parameters to configure the service network slice resource parameters.

[0054] Furthermore, the traffic management module 14 is used to perform the following methods: Real-time monitoring of service traffic distribution is performed using the service network slice resource parameters to obtain slice resource status feedback parameters; the service network slice resource parameters are optimized and updated based on the slice resource status feedback parameters, and closed-loop management of service traffic distribution is performed using the updated service network slice resource parameters.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A CPE service offloading management method based on 5G network slicing, characterized in that, The method includes: The application layer identification data and network layer feature data are collected by the CPE device, and the service type is identified by the application layer identification data and network layer feature data using a service classifier to obtain CPE service type information. Acquire 5G network slice resources, dynamically allocate the 5G network slice resources based on the CPE service type information, and determine the network slice resource allocation parameters; Establish a business traffic prediction channel, use the business traffic prediction channel to perform traffic monitoring and prediction on the CPE business type information, and output CPE business traffic prediction parameters. The service traffic allocation engine optimizes the network slice resource allocation parameters based on the CPE service traffic prediction parameters, configures the service network slice resource parameters, and performs closed-loop management of service traffic allocation through the service network slice resource parameters.

2. The CPE service offloading management method based on 5G network slicing as described in claim 1, characterized in that, Obtain CPE service type information, including: The application layer identification data and network layer feature data are subjected to invalid data filtering, duplicate data removal and timestamp standardization to obtain standard application layer identification data and standard network layer feature data. Numerical, categorical, and temporal features are extracted from the standard application layer identifier data and standard network layer feature data to obtain a key feature set of business application networks. A business classifier is constructed by training a classification system based on the historical CPE business dataset. The service classifier is used to identify the service type of the key feature set of the service application network to obtain CPE service type information.

3. The CPE service offloading management method based on 5G network slicing as described in claim 2, characterized in that, Building a business classifier includes: The CPE service history dataset is characterized by feature identification and service type labeling to obtain a CPE service type sample set; The information gain of each sample in the CPE business type sample set is calculated and recursively split using a decision tree to construct an initial classification model. The initial classification model is subjected to node backtracking and error evaluation pruning to obtain a post-processed classification model. The performance of the post-processing classification model is verified and optimized to minimize the model loss function, thereby obtaining the business classifier.

4. The CPE service offloading management method based on 5G network slicing as described in claim 1, characterized in that, Determine network slice resource allocation parameters, including: Based on the CPE service scenario requirements, the 5G network slice resources are analyzed for slice configuration, and a service scenario type-slice resource configuration parameter template library is generated. Based on the CPE service type information, the initial slice resource parameter template is obtained by matching and selecting with the service scenario type-slice resource configuration parameter template library. The SDN controller monitors the slice resource status parameters in real time. If the slice resource status parameters are lower than the preset resource threshold, a dynamic adjustment mechanism is triggered. Based on the aforementioned dynamic adjustment mechanism, the initial slice resource parameter template is dynamically allocated to determine the network slice resource allocation parameters.

5. The CPE service offloading management method based on 5G network slicing as described in claim 4, characterized in that, Based on the aforementioned dynamic adjustment mechanism, the initial slice resource parameter template is dynamically allocated to determine the network slice resource allocation parameters, including: Based on the aforementioned dynamic adjustment mechanism, a multi-dimensional business priority evaluation system is obtained; The priority assessment of each business type in the CPE business type information is performed according to the multi-dimensional business priority assessment system to generate a business type priority assessment matrix. The initial slice resource parameter template is dynamically adjusted and allocated based on the service type priority evaluation matrix to determine the network slice resource allocation parameters.

6. The CPE service offloading management method based on 5G network slicing as described in claim 5, characterized in that, Based on the service type priority evaluation matrix, the initial slice resource parameter template is dynamically adjusted and allocated to determine the network slice resource allocation parameters, including: Based on the business type priority evaluation matrix, the business types are sorted to obtain a business type priority sequence; Based on the priority sequence of the business types, determine the business type diversion decision ratio; Based on the traffic diversion decision ratio of the service type, the initial slice resource parameter template is dynamically adjusted and allocated to determine the network slice resource allocation parameters.

7. The CPE service offloading management method based on 5G network slicing as described in claim 1, characterized in that, Establish a business traffic prediction channel, including: Collect historical data of business traffic, clean the historical data of business traffic and timestamp it to obtain a time-series data of business traffic. Obtain a set of factors influencing business traffic, and label the time-series dataset of business traffic according to the set of factors influencing business traffic and a preset time window to obtain a time-series sample set of business traffic. A time series network is used to train traffic prediction on the time series sample set of the service traffic, and a service traffic prediction channel is established.

8. The CPE service offloading management method based on 5G network slicing as described in claim 1, characterized in that, Configure service network slice resource parameters, including: According to the business traffic splitting strategy engine, the slice resource splitting strategy library is invoked. The slice resource splitting strategy library includes priority splitting rules, peak period splitting rules, and backup slice splitting rules. Based on the CPE service traffic prediction parameters, the resource gap is calculated to obtain the CPE service traffic resource gap parameters; According to the slice resource allocation strategy library, the network slice resource allocation parameters are optimized based on the CPE service traffic resource gap parameters, and the service network slice resource parameters are configured.

9. The CPE service offloading management method based on 5G network slicing as described in claim 1, characterized in that, Service traffic splitting closed-loop management is performed using the service network slice resource parameters, including: Real-time monitoring of service diversion is performed using the service network slice resource parameters to obtain slice resource status feedback parameters. The service network slice resource parameters are optimized and updated based on the slice resource status feedback parameters, and the updated service network slice resource parameters are used for closed-loop management of service diversion.

10. A CPE service offloading management system based on 5G network slicing, characterized in that, The system is used to implement the CPE service offloading management method based on 5G network slicing as described in any one of claims 1-9, the system comprising: Data acquisition module: Collects application layer identification data and network layer feature data through CPE device, and uses a service classifier to identify the service type of the application layer identification data and network layer feature data to obtain CPE service type information; Resource allocation module: acquires 5G network slice resources, dynamically allocates the 5G network slice resources based on the CPE service type information, and determines the network slice resource allocation parameters; Traffic prediction module: Establishes a business traffic prediction channel, uses the business traffic prediction channel to perform traffic monitoring and prediction on the CPE business type information, and outputs CPE business traffic prediction parameters; Traffic splitting management module: Based on the CPE service traffic prediction parameters, the traffic splitting strategy engine optimizes the network slice resource allocation parameters, configures the service network slice resource parameters, and performs closed-loop management of service splitting through the service network slice resource parameters.

Images