An adjustment method and system for 5G network slicing

Abstract

The present application discloses an adjustment method and system for 5G network slicing, which relates to the field of 5G networks. The method includes: obtaining historical demand information of a target service; determining an initial network slicing policy based on the historical demand information; determining an initial network slice based on the initial network slicing policy; monitoring performance data of each of the initial network slices; obtaining priority information of the target service corresponding to each network slice; and making adjustments to the initial network slices based on the performance data and the priority information to determine optimized network slices.

Description

Technical Field

[0001] This specification relates to the field of 5G networks. More specifically, this application relates to a method and system for adjusting 5G network slices. Background Art

[0002] With the rapid development of 5G technology, network slicing, as a key technology, allows operators to create multiple independent virtual networks on a shared physical network infrastructure. Each network slice can independently allocate resources and set quality of service (QoS) parameters according to different service requirements. This flexible network resource management method greatly improves the network utilization rate and service quality. However, existing network slice management solutions usually rely on preset policies and cannot flexibly respond to dynamic network environments and service requirements, resulting in low resource utilization efficiency and an inability to effectively guarantee the priority of critical services. Summary of the Invention

[0003] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further elaborated in the Detailed Description section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0004] In a first aspect, this application proposes a method for adjusting 5G network slices, including:

[0005] Obtaining historical demand information of a target service;

[0006] Determining an initial network slice policy based on the above historical demand information;

[0007] Determining an initial network slice based on the above initial network slice policy;

[0008] Monitoring performance data of each of the above initial network slices;

[0009] Obtaining priority information of the target service corresponding to each network slice;

[0010] Adjusting the above initial network slices based on the above performance data and the above priority information to determine an optimized network slice.

[0011] In a feasible implementation, the above determining an initial network slice policy based on the above historical demand information includes:

[0012] Based on the above historical demand information, a preset network slicing strategy is obtained using a genetic algorithm. Among them, the above historical demand information includes bandwidth usage information, latency requirement information, and connection density information of the target service. The above bandwidth usage information is obtained from network logs, the above latency requirement information is obtained from user usage records, and the above connection density information is obtained from service load data;

[0013] An initial network slicing strategy is determined according to the above preset network slicing strategy through a reinforcement learning algorithm.

[0014] In a feasible implementation manner, the obtaining of the preset network slicing strategy using a genetic algorithm based on the above historical demand information includes:

[0015] Define the encoding method of the network slicing strategy, and encode the bandwidth usage information, latency requirement information, and connection density information of each above target service, which is represented as genes in the chromosome;

[0016] Randomly extract typical resource allocation patterns from the historical demand information and randomly generate an initial population;

[0017] Design a fitness function to evaluate the quality of each chromosome;

[0018] Based on the above fitness function, selection operations, crossover operations, and mutation operations are performed, and iterative operations are continuously carried out until the fitness function improvement rate is less than a preset threshold or the number of iterations reaches a preset number, and the above preset network slicing strategy is output.

[0019] In a feasible implementation manner, the determination of the initial network slicing strategy according to the above preset network slicing strategy through a reinforcement learning algorithm includes:

[0020] Model the network environment as the environmental state in reinforcement learning, and model the above preset network slicing strategy as the behavior of the intelligent agent. Among them, the above environmental state includes the current network state, resource utilization, and service demand:

[0021] Design a reward function to evaluate the quality of each action;

[0022] Adopt the Q-learning reinforcement learning algorithm, with the goal of maximizing the cumulative reward, and update the strategy by repeatedly interacting with the environment;

[0023] At each time step, the intelligent agent observes the current state, selects an action to adjust the network slicing strategy, and updates the network slicing strategy according to the above reward function;

[0024] In the case of reaching the preset number of iterations and / or policy convergence, the above initial network slicing strategy is output.

[0025] In a feasible implementation manner, adjusting the above initial network slice based on the above performance data and the above priority information to determine an optimized network slice includes:

[0026] Defining a plurality of objective functions according to the above performance data and the above priority information, wherein the above performance data includes latency data, bandwidth utilization data, energy consumption data, and quality of service data; the above plurality of objective functions include a latency objective function, a bandwidth utilization objective function, an energy consumption objective function, and a quality of service objective function, and each objective function includes a priority information, and the above priority information is used to adjust the weight of the objective function;

[0027] Using the NSGA-II algorithm to perform multi-objective optimization on all the objective functions to adjust the above initial network slice and determine an optimized network slice, wherein the above optimized network slice includes a resource allocation slice, a quality of service slice, a bandwidth slice, a latency optimization slice, and an energy consumption optimization slice.

[0028] In a feasible implementation manner, the above optimized network slice further includes a resource isolation slice and a resource protection slice.

[0029] The above method further includes:

[0030] Obtaining performance bottleneck information and exception information;

[0031] Determining a resource isolation slice and a resource protection slice based on the above performance bottleneck information, the above exception information, and the above priority information.

[0032] In a feasible implementation manner, determining a resource isolation slice and a resource protection slice based on the above performance bottleneck information, the above exception information, and the above priority information includes:

[0033] Analyzing the above performance bottleneck information and the above exception information based on a hierarchical clustering algorithm to obtain a plurality of candidate areas for categories;

[0034] Calculating the priority score of each candidate area for a category according to the priority information corresponding to all target services in each above candidate area for a category;

[0035] Determining a resource isolation slice and a resource protection slice based on the priority score of each candidate area for a category and a priority score threshold.

[0036] In a feasible implementation manner, calculating the priority score of each candidate area for a category according to the priority information corresponding to all target services in each above candidate area for a category includes:

[0037] Calculating the priority score S(Ci) according to the following formula:

[0038]

[0039] Among them, p ij represents the priority of the target service, and w ij is the weight coefficient corresponding to the j-th target service in the j-th target service selection area of the C i -th category. i is the index of the category target service selection area, and j is the index of a certain target service in the category target service selection area.

[0040] In a feasible implementation manner, the specific steps for determining the weight coefficients corresponding to the target services in each category target service selection area include:

[0041] Determine the basic weight coefficients of service importance corresponding to the target services in each category target service selection area based on the number of service users, business income contribution, and business continuity requirements of each target service in each category target service selection area;

[0042] Determine the basic weight coefficients of resource occupancy corresponding to the target services in each category target service selection area based on the bandwidth information, CPU utilization information, and storage space information of each target service in each category target service selection area;

[0043] Determine the weight coefficients corresponding to the target services in each category target service selection area based on the above basic weight coefficients of service importance, the above basic weight coefficients of resource occupancy, the first proportion coefficient, and the second proportion coefficient corresponding to the target services in each category target service selection area. Among them, the above first proportion coefficient is the proportion coefficient corresponding to the basic weight coefficient of service importance, and the above second proportion coefficient is the proportion coefficient corresponding to the basic weight coefficient of resource occupancy.

[0044] Second aspect, an adjustment system for a 5G network slice of the present application includes:

[0045] A first acquisition unit, configured to acquire historical demand information of a target service;

[0046] A first determination unit, configured to determine an initial network slice policy based on the above historical demand information;

[0047] A second determination unit, configured to determine an initial network slice based on the above initial network slice policy;

[0048] A monitoring unit, configured to monitor the performance data of each of the above initial network slices;

[0049] A second acquisition unit, configured to acquire priority information of the target service corresponding to each network slice;

[0050] An adjustment unit, configured to adjust the above initial network slice based on the above performance data and the above priority information to determine an optimized network slice.

[0051] In summary, the adjustment method for 5G network slicing according to the embodiments of the present application includes: obtaining historical demand information of a target service; determining an initial network slicing strategy based on the above historical demand information; determining an initial network slice based on the above initial network slicing strategy; monitoring performance data of each of the above initial network slices; obtaining priority information of the target service corresponding to each network slice; and making adjustments to the above initial network slices based on the above performance data and the above priority information to determine an optimized network slice. In the prior art, network slicing strategies are usually based on static configurations and are difficult to cope with dynamic network environments. By monitoring the performance data of network slices in real time and combining the priority information of target services, the present application can dynamically adjust network slicing strategies, making network resource allocation more flexible and enabling quick response to changes in network conditions. In existing solutions, preset network slicing strategies may lead to over-allocation or under-allocation of resources, affecting the overall efficiency of the network. By analyzing the historical demand information of target services, formulating an initial network slicing strategy, and optimizing it based on real-time data, the present application ensures that resources can be allocated and used more reasonably, maximizing the utilization rate of network resources. The present application particularly emphasizes the priority management of target services. After obtaining the service priority information corresponding to each network slice, the system can prioritize the performance requirements of high-priority services under limited resources. This is of great significance for ensuring the service quality of critical services, especially in the case of concurrent operation of multiple services, effectively avoiding service performance degradation caused by resource contention. The present application is applicable to various complex network environments and diverse service requirements. Whether in a mobile network scenario with frequent fluctuations in user demands or in an enterprise private network with strict service quality requirements, this solution can provide an effective network slicing adjustment method to ensure network stability and service quality. Through these improvements, the present application can better meet the diverse and complex service requirements in 5G networks, achieve efficient management of network resources and stable guarantee of high-quality services, and has significant technological progress and practical application value.

[0052] For the adjustment method for 5G network slicing proposed in the present application, other advantages, objectives, and features of the present application will be partially reflected by the following description and partially understood by those skilled in the art through research and practice of the present application. Brief Description of the Drawings

[0053] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of illustrating the preferred embodiments and are not considered to be a limitation of this specification. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0054] Figure 1A schematic flowchart of an adjustment method for a 5G network slice provided by an embodiment of the present application;

[0055] Figure 2 A video transmission system for live vocal music teaching provided by an embodiment of the present application. Detailed implementation manners

[0056] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0057] Please refer to Figure 1 , which is a schematic flowchart of an adjustment method for a 5G network slice provided by an embodiment of the present application, and includes:

[0058] S110. Obtain historical demand information of the target service;

[0059] S120. Determine an initial network slice strategy based on the above historical demand information;

[0060] S130. Determine an initial network slice based on the above initial network slice strategy;

[0061] S140. Monitor the performance data of each of the above initial network slices;

[0062] S150. Obtain the priority information of the target service corresponding to each network slice;

[0063] S160. Make adjustments to the above initial network slices based on the above performance data and the above priority information to determine an optimized network slice.

[0064] Exemplarily, the system collects historical demand information related to the target service. Such information includes, but is not limited to, bandwidth usage information, latency requirement information, connection density information, etc. of the target service in the past period of time. These data provide a basis for formulating the network slice strategy in the subsequent steps.

[0065] After collecting historical requirement information, the system determines an initial network slicing strategy based on this information. This strategy will define how to allocate resources for target services in the network to ensure service quality and efficiency.

[0066] According to the previously determined initial network slicing strategy, the system generates specific initial network slices. Each network slice corresponds to a certain resource allocation method to ensure that specific target services can obtain the resources that meet their requirements.

[0067] After the initial network slice configuration is completed, the system monitors the performance data of each network slice in real time. These performance data may include bandwidth utilization rate, latency, packet loss rate, etc., aiming to evaluate the actual operation effect of the network slice.

[0068] The system also obtains the priority information of the target services corresponding to each network slice. The priority information determines the resource allocation order and importance among different services. For example, some critical services may have higher priorities and need to ensure their performance first.

[0069] Finally, the system adjusts the initial network slices according to the collected performance data and priority information. These adjustments aim to optimize the allocation of network resources to ensure that high-priority target services can obtain sufficient resources in the network, while improving the efficiency and performance of the overall network slices. The optimized network slice configuration is finally determined.

[0070] In summary, in the prior art, network slicing strategies are usually based on static configurations and are difficult to cope with dynamic network environments. By monitoring the performance data of network slices in real time and combining the priority information of target services, this application can dynamically adjust network slicing strategies, making network resource allocation more flexible and enabling rapid response to changes in network conditions. In existing solutions, preset network slicing strategies may lead to over-allocation or under-allocation of resources, affecting the overall efficiency of the network. By analyzing the historical demand information of target services, this application formulates an initial network slicing strategy and optimizes it based on real-time data to ensure that resources can be allocated and used more reasonably, maximizing the utilization rate of network resources. This application pays particular attention to the priority management of target services. After obtaining the service priority information corresponding to each network slice, the system can prioritize the performance requirements of high-priority services in the case of limited resources. This is of great significance for ensuring the quality of service of critical services, especially in the case of concurrent operation of multiple services, which can effectively avoid the degradation of service performance caused by resource contention. This application is applicable to various complex network environments and diverse service requirements. Whether in a mobile network scenario with frequent fluctuations in user demands or in an enterprise private network with strict service quality requirements, this solution can provide an effective method for network slice adjustment to ensure network stability and service quality. Through these improvements, this application can better meet the diverse and complex service requirements in 5G networks, achieve efficient management of network resources and stable guarantee of high-quality services, and has significant technological progress and practical application value.

[0071] In some examples, determining the initial network slicing strategy based on the above historical demand information includes:

[0072] Using a genetic algorithm to obtain a preset network slicing strategy based on the above historical demand information. Among them, the above historical demand information includes bandwidth usage information, delay requirement information, and connection density information of the target service. The above bandwidth usage information is obtained from network logs, the above delay requirement information is obtained from user usage records, and the above connection density information is obtained from business load data;

[0073] Determining the initial network slicing strategy according to the preset network slicing strategy through a reinforcement learning algorithm.

[0074] Exemplarily, the historical demand information is obtained from multiple data sources, including network logs, user usage records, business load data, etc. These data sources provide key performance indicator information of the target service, such as bandwidth usage information, delay requirement information, and connection density information. These information describe the network resource requirements of the target service in different time periods and provide data support for subsequent steps.

[0075] After collecting historical requirement information, the system uses a genetic algorithm based on this information to obtain a preset network slicing strategy. This strategy will pre-define how to allocate resources for target services in the network to approach the best quality of service as much as possible. Based on the preset network slicing strategy obtained through the genetic algorithm as described above, the system further determines an initial network slicing strategy through a reinforcement learning algorithm.

[0076] In the embodiments of the present application, by combining a genetic algorithm and a reinforcement learning algorithm, a network slicing strategy formulation process is formed that can not only consider historical requirement information but also continuously optimize in a dynamic environment. This method can achieve more efficient resource management and better service guarantee in a complex 5G network.

[0077] In some examples, obtaining the preset network slicing strategy using the genetic algorithm based on the above historical requirement information includes:

[0078] Define the encoding method of the network slicing strategy, and encode the bandwidth usage information, latency requirement information, and connection density information of each of the above target services, which is represented as genes in a chromosome;

[0079] Randomly extract typical resource allocation patterns from the historical requirement information and randomly generate an initial population;

[0080] Design a fitness function to evaluate the quality of each chromosome;

[0081] Based on the above fitness function, perform selection operations, crossover operations, and mutation operations, and continuously perform iterative operations until the fitness function improvement rate is less than a preset threshold or the number of iterations reaches a preset number, and output the above preset network slicing strategy.

[0082] Exemplarily, before starting the optimization, it is first necessary to define the encoding method of the network slicing strategy. In this solution, the encoding method is to represent the bandwidth usage information, latency requirement information, and connection density information of each target service as genes in a chromosome. Each gene represents a specific resource allocation strategy, and the entire chromosome represents a complete network slicing strategy.

[0083] To start the optimization process of the genetic algorithm, first randomly extract typical resource allocation patterns from the historical requirement information and randomly generate an initial population according to these patterns. Each individual in this initial population is a possible network slicing strategy, generated randomly to ensure the diversity of the population, so as to have more possibilities to find the optimal solution in subsequent iterations.

[0084] To evaluate the quality of each individual (i.e., each network slicing strategy), a fitness function is designed. This fitness function is used to measure the effect of the strategy, specifically expressed as:

[0085] F(x) = α·U(x) + β·Q(x) - γ·C(x)

[0086] Wherein, U(x) represents the resource utilization rate, Q(x) represents the quality of service, C(x) represents the cost, and α, β, and γ are weight parameters; these three indicators jointly determine the quality of the network slicing strategy. The higher the resource utilization rate, the better the quality of service, and the lower the cost of the strategy, the higher its fitness.

[0087] In each iteration, based on the fitness function, selection operations are performed on the individuals in the population, and individuals with higher fitness are preferentially retained. Then, two chromosomes (i.e., two network slicing strategies) are combined through crossover operations to generate new individuals. At the same time, mutation operations are applied to randomly modify some gene values to increase the diversity of the population and prevent falling into local optimal solutions.

[0088] The genetic algorithm continuously performs selection, crossover, and mutation operations, and gradually optimizes the individuals in the population through iterative operations. When the improvement rate of the fitness function is less than the preset threshold, or the number of iterations reaches the preset number of times, the algorithm terminates and outputs the final preset network slicing strategy.

[0089] Through the method proposed in this embodiment, the system can extract an optimized preset network slicing strategy from historical demand information, laying a foundation for the formulation of the subsequent initial network slicing strategy. This method ensures the efficient utilization of network resources, improves the quality of service, and provides a better network slicing solution while meeting certain cost constraints.

[0090] In some examples, determining the initial network slicing strategy according to the above preset network slicing strategy through the reinforcement learning algorithm includes:

[0091] Model the network environment as the environmental state in reinforcement learning, and model the above preset network slicing strategy as the behavior of the agent, wherein the above environmental state includes the current network state, resource utilization, and service demand:

[0092] Design a reward function to evaluate the quality of each action;

[0093] Adopt the Q-learning reinforcement learning algorithm, with the goal of maximizing the cumulative reward, and update the strategy by repeatedly interacting with the environment;

[0094] At each time step, the agent observes the current state, selects an action to adjust the network slicing strategy, and updates the network slicing strategy according to the above reward function;

[0095] When the preset number of iterations and / or strategy convergence is reached, output the above initial network slicing strategy.

[0096] Exemplarily, the network environment is modeled as the environmental state in reinforcement learning. Specifically, the environmental state includes the current network state, resource utilization, and service demand. These state information provide the background in which the current network slicing policy is located, enabling the agent to make reasonable decisions based on this information.

[0097] The above-mentioned preset network slicing policy is modeled as the agent's behavior. The agent's behavior is to select a specific network slicing policy or adjust the existing policy under the given environmental state. Through this modeling, each action of the agent corresponds to an adjustment of the network slicing policy.

[0098] Design a reward function to evaluate the quality of each action. The above reward function is:

[0099] R(s, a) = δ · Q(s, a) - ∈ · D(s, a)

[0100] where Q(s, a) represents the improvement in service quality obtained after taking action a in state s, D(s, a) represents the cost of resource usage, and δ, ∈ are the corresponding weight information; the goal of the reward function is to maximize the improvement in service quality while minimizing the cost of resource usage.

[0101] Adopt the Q-learning reinforcement learning algorithm. By continuously interacting with the environment, update the policy to maximize the cumulative reward. At each time step, the agent observes the current state and selects an optimal action to adjust the network slicing policy. The agent will update its policy using the Q-learning algorithm according to the reward obtained after executing the action, so that the policy is gradually optimized and can better cope with different network states and service demands.

[0102] At each time step, the agent observes the current state and selects an action according to the Q-learning algorithm. This action corresponds to an adjustment of the network slicing policy. Then, the agent calculates the quality of this action according to the above reward function and uses this information to update the policy. Through continuous policy updates, the agent can learn which network slicing policies can bring the maximum cumulative reward under different environmental states.

[0103] When the preset number of iterations and / or policy convergence is reached, the algorithm stops and outputs the final initial network slicing policy. At this time, the policy has been iterated and optimized multiple times and can better balance the relationship between service quality improvement and resource usage cost.

[0104] For the method provided in this embodiment, the system can continuously optimize the network slicing strategy based on the reinforcement learning algorithm in a dynamic and complex network environment, thereby effectively improving the utilization rate of network resources and service quality. The finally determined initial network slicing strategy can adapt to diverse service requirements and perform excellently in various environments.

[0105] In some examples, adjusting the above initial network slice based on the above performance data and the above priority information to determine an optimized network slice includes:

[0106] Defining multiple objective functions according to the above performance data and the above priority information, where the above performance data includes latency data, bandwidth utilization data, energy consumption data, and service quality data; the above multiple objective functions include a latency objective function, a bandwidth utilization objective function, an energy consumption objective function, and a service quality objective function, and each objective function includes a priority information, which is used to adjust the weight of the objective function;

[0107] Using the NSGA-II algorithm to perform multi-objective optimization on all the objective functions to adjust the above initial network slice and determine an optimized network slice, where the above optimized network slice includes a resource allocation slice, a service quality slice, a bandwidth slice, a latency optimization slice, and an energy consumption optimization slice.

[0108] Exemplarily, the performance data includes:

[0109] Latency data: Represents the response time of services in the network slice, which directly affects the user experience.

[0110] Bandwidth utilization data: Represents the actual usage of network bandwidth, as a percentage of the available bandwidth.

[0111] Energy consumption data: Represents the energy consumed during the operation of the network slice, which is related to the energy efficiency of the network.

[0112] Service quality data: Represents the overall quality of services provided by the network slice to users, which may include multiple metrics such as packet loss rate and jitter.

[0113] During the process of adjusting the initial network slice, multiple objective functions are defined according to the above latency data, bandwidth utilization data, energy consumption data, service quality data, and priority information. Each objective function corresponds to one of the above performance data and includes a priority information for adjusting the weight of the objective function.

[0114] Latency objective function: Reflects the impact of network latency on the overall performance, and the priority information is used for weighted calculation so that services with high priority receive more attention in terms of latency.

[0115] Bandwidth utilization objective function: Reflects the impact of bandwidth utilization on network resources. Priority information is used to adjust the weight of bandwidth utilization to ensure the bandwidth requirements of high-priority services.

[0116] Energy consumption objective function: Reflects the energy consumption situation during network operation. Priority information is used to adjust the weight of the energy consumption objective to ensure that energy consumption is optimized in the case of high-priority services.

[0117] Quality of service objective function: Reflects the impact of network service quality on user experience. Priority information is used to adjust the weight of service quality to ensure that the quality requirements of high-priority services are met.

[0118] After defining the objective functions, the NSGA-II algorithm is used to perform multi-objective optimization on all objective functions. NSGA-II is an advanced evolutionary algorithm that can effectively handle multiple conflicting objectives and generate a set of Pareto optimal solutions.

[0119] Initial population generation: Generate an initial population containing multiple candidate network slicing strategies. Each strategy is an individual representing a possible network slicing adjustment plan.

[0120] Fitness evaluation: For each individual in the population, perform fitness evaluation according to the defined objective functions. Since each objective function contains priority information, more optimization weight will be automatically given to high-priority services during the evaluation process.

[0121] Non-dominated sorting and selection: NSGA-II sorts individuals through non-dominated sorting and retains those individuals that perform well in each objective function.

[0122] Crossover and mutation operations: Generate new individuals through crossover and mutation operations to maintain the diversity of the population and explore new network slicing strategies.

[0123] Iterative optimization: Repeatedly perform fitness evaluation, non-dominated sorting, selection, crossover, and mutation operations until the preset number of iterations or convergence conditions are reached.

[0124] Resource allocation slice: Used to manage the dynamic allocation of network resources and adjust according to real-time demands and performance data.

[0125] Quality of service slice (QoS slice): Ensures that the quality requirements of specific services are met and provides different resource guarantees for services with different priorities.

[0126] Bandwidth slice: Used to guarantee the bandwidth requirements between different applications and adjust according to the network load situation.

[0127] Latency Optimization Slice: Specifically designed for low-latency services, such as application scenarios that require extremely low latency like autonomous driving and telemedicine.

[0128] Energy Consumption Optimization Slice: A slice designed for energy-saving optimization, suitable for network scenarios sensitive to energy consumption.

[0129] Finally, a set of Pareto optimal solutions generated by the NSGA-II algorithm represents different optimized network slicing strategies. These strategies make trade-offs between different objective functions, considering multiple performance metrics such as latency, bandwidth utilization, energy consumption, and quality of service, while fully considering the impact of priority information. The system can select an optimal solution according to specific requirements as the final optimized network slice.

[0130] In some examples, the above-mentioned optimized network slices also include resource isolation slices and resource protection slices.

[0131] The above method further includes:

[0132] Obtaining performance bottleneck information and exception information;

[0133] Determining resource isolation slices and resource protection slices based on the above performance bottleneck information, the above exception information, and the above priority information.

[0134] Exemplarily, the optimized network slice not only includes general network slicing strategies but also resource isolation slices and resource protection slices. The method determines resource isolation slices and resource protection slices by obtaining performance bottleneck information and exception information and combining priority information, thereby further optimizing the allocation and utilization of network resources.

[0135] First, by monitoring the network operation status, performance bottleneck information and exception information are obtained. These information include but are not limited to:

[0136] Performance bottleneck information: Reflects areas in the network where there are resource shortages or service quality degradation. For example, insufficient bandwidth, excessive latency, or overloaded computing resources, etc.

[0137] Exception information: Reflects abnormal events or emergencies that occur in the network, such as network congestion, link failures, or service interruptions, etc.

[0138] After obtaining the performance bottleneck information and exception information, and combining the priority information, the system begins to determine resource isolation slices and resource protection slices. The priority information indicates the importance of different services, ensuring that high-priority services are given priority in resource allocation.

[0139] Resource isolation slice: The resource isolation slice is used to isolate the resource areas affected by performance bottleneck information and exception information, preventing these problems from spreading to other network slices and affecting the overall network performance. By analyzing the performance bottleneck information and exception information, it is determined which resource areas may need to be isolated based on this information. Combining with the priority information, if some low-priority services occupy the resources that high-priority services may need, these low-priority services are isolated through the resource isolation slice to ensure the resource requirements of high-priority services.

[0140] The resource protection slice is used to ensure that high-priority services can obtain sufficient resource support and maintain service quality in case of resource shortage or exceptions. Combining with the priority information, the resource requirements corresponding to high-priority services are identified, and resource protection slices are established within possible resource areas according to these requirements. The resource protection slice ensures that when performance bottleneck information and exception information occur, the resources of high-priority services will not be occupied or interfered by low-priority services.

[0141] In some examples, determining the resource isolation slice and the resource protection slice based on the above performance bottleneck information, the above exception information, and the above priority information includes:

[0142] Performing an analysis operation on the above performance bottleneck information and the above exception information based on the hierarchical clustering algorithm to obtain multiple candidate areas for categories;

[0143] Calculating the priority score of each candidate area for a category according to the priority information corresponding to all target services in each of the above candidate areas for categories;

[0144] Determining the resource isolation slice and the resource protection slice based on the priority score of each candidate area for a category and the priority score threshold.

[0145] Exemplarily, performing an analysis operation on the obtained performance bottleneck information and exception information using the hierarchical clustering algorithm. The hierarchical clustering algorithm is an unsupervised learning algorithm that gradually aggregates similar data points into clusters by calculating the similarity or distance between data points.

[0146] Taking the performance bottleneck information and exception information as input data to construct a data matrix. Each data point represents a specific performance bottleneck or exception situation, and the similarity between data points is calculated based on this information.

[0147] Using the hierarchical clustering algorithm to gradually aggregate similar performance bottleneck information and exception information into clusters. Through the clustering operation of the algorithm, the system can identify multiple candidate areas for categories, and each candidate area for a category represents a group of resource areas with similar performance bottlenecks and exception situations.

[0148] After obtaining multiple categories of candidate areas, the system will calculate the priority score of each category of candidate areas according to the priority information corresponding to all target services in each category of candidate areas. The specific steps are as follows:

[0149] Priority information integration: Collect the priority information of all target services included in each category of candidate areas. The priority information of each target service represents the importance of the service, which is usually predefined by the service provider or network operator.

[0150] Priority score calculation: According to the formula, perform a weighted sum of the priority information p ij of all target services in each category of candidate areas and the corresponding weight coefficient w ij to calculate the priority score S(C i ). The priority score reflects the overall importance of this category of candidate areas and the urgency of resource requirements.

[0151] After calculating the priority score of each category of candidate areas, the system will determine resource isolation slices and resource protection slices based on these scores and a preset priority score threshold. Compare the priority score of each category of candidate areas with the preset priority score threshold. This threshold is used to distinguish high-priority areas and low-priority areas. For a category of candidate areas with a priority score lower than the threshold, the system will identify it as a resource isolation slice. The resources in these slices will be isolated to prevent them from having a negative impact on the overall network performance, especially in the case of performance bottlenecks or anomalies. For a category of candidate areas with a priority score higher than the threshold, the system will identify it as a resource protection slice. The resources in these slices will be preferentially protected and allocated to ensure that high-priority services can still maintain good performance and service quality when resources are scarce.

[0152] In some examples, the above calculation of the priority score of each category of candidate areas according to the priority information corresponding to all target services in each category of candidate areas includes:

[0153] Calculate the priority score S(Ci) according to the following formula:

[0154]

[0155] where p ij represents the priority of the target service, w ij is the weight coefficient corresponding to the jth target service in the C i th category of candidate areas, i is the index of the category of candidate areas, and j is the index of a certain target service in this category of candidate areas.

[0156] Exemplarily, P ij represents the category of candidate areas C iThe priority of target service j. The priority reflects the importance of the target service in the entire network and is usually determined by preset business requirements or policies. High-priority target services are crucial for ensuring service quality and user experience.

[0157] w ij Represents the weight coefficient of target service j in the candidate category area Ci. This weight coefficient reflects the proportion or importance of target service j in this candidate category area. The weight coefficient is usually determined by the resource requirements of the service, the number of service users, or other key indicators. It helps to further refine the importance assessment of the target service for the entire candidate category area.

[0158] The priority score S(C i ) is obtained by weighted summation of the priorities P i of all target services within the candidate category area C ij and their corresponding weight coefficients w ij . Specifically, the priority P ij of each target service is multiplied by its corresponding weight coefficient w ij . The resulting value represents the importance contribution of this target service to the candidate category area. Then, the contribution values of all target services are added together to obtain the overall priority score S(C i ).

[0159] The calculated priority score S(C i ) represents the overall priority of the candidate category area C i . The higher the score, the more high-priority services are included in this candidate category area, or the greater the demand of these services for network resources. Therefore, the priority score is an important basis for determining resource isolation areas and resource protection areas. Based on the score, the system can reasonably allocate resources to ensure that high-priority target services are preferentially guaranteed when resources are scarce, while low-priority services may be isolated or restricted to avoid affecting critical services.

[0160] In some examples, the specific steps for determining the weight coefficient of the above target service in the candidate category area include:

[0161] Determine the basic weight coefficient of service importance based on the number of service users, business revenue contribution, and business continuity requirements of the target service;

[0162] Determine the basic weight coefficient of resource occupancy based on the bandwidth information, CPU utilization information, and storage space information of the target service;

[0163] Determine the above-mentioned basic weight coefficient of resource occupancy based on the above-mentioned basic weight coefficient of service importance, the above-mentioned basic weight coefficient of resource occupancy, the first proportion coefficient, and the second proportion coefficient, where the above-mentioned first proportion coefficient is the proportion coefficient corresponding to the above-mentioned basic weight coefficient of service importance, and the above-mentioned second proportion coefficient is the proportion coefficient corresponding to the above-mentioned basic weight coefficient of resource occupancy.

[0164] Exemplarily, determine the basic weight coefficient of service importance by analyzing three key indicators of the target service:

[0165] Number of service users: The number of users of a service directly reflects the user coverage of the service. The more users, the higher the importance of the service.

[0166] Business revenue contribution: It refers to the contribution ratio of the service to the overall business revenue. Services with a high business revenue contribution are usually regarded as strategic key businesses, so their importance is higher.

[0167] Business continuity requirement: Some services may have extremely high requirements for business continuity, such as critical tasks or real-time services. The higher the business continuity requirement, the greater the importance of the service.

[0168] By comprehensively considering these three indicators, calculate the basic weight coefficient of service importance, which reflects the relative importance of the target service in the entire business system.

[0169] It can be determined by the following formula:

[0170]

[0171] Wi mportance : The basic weight coefficient of service importance, reflecting the relative importance of the target service in the entire business system.

[0172] U i： The number of users of the target service.

[0173] R i : The business revenue contribution of the target service.

[0174] C i : The business continuity requirement of the target service.

[0175] U max : The maximum value of the number of users among all services, used for normalization.

[0176] R max : The maximum value of the business revenue contribution among all services, used for normalization.

[0177] C max : The maximum value of the business continuity requirement among all services, used for normalization.

[0178] α, β, and γ are the weight coefficients for the number of users, business revenue contribution, and business continuity requirements respectively, which are used to adjust the influence degree of each indicator on the overall weight. These weights can be set according to actual business needs to reflect the importance of each indicator.

[0179] Next, determine the basic weight coefficient of resource occupancy based on the bandwidth information, CPU utilization information, and storage space information of the target service:

[0180] Bandwidth information: It represents the amount of bandwidth occupied by the target service in the network. The greater the bandwidth occupancy, the higher the resource demand.

[0181] CPU utilization information: It reflects the occupancy of computing resources (CPU) by the target service during the processing. High CPU utilization indicates that the service requires more computing resources.

[0182] Storage space information: The size of the storage space required by the service is also an important factor in measuring its resource occupancy. The greater the storage space demand, the higher the resource occupancy.

[0183] Based on the above resource occupancy indicators, calculate the basic weight coefficient of resource occupancy, which quantifies the demand intensity of the target service for network and computing resources.

[0184] Finally, combine the basic weight coefficient of service importance and the basic weight coefficient of resource occupancy obtained previously, and apply two proportion coefficients to determine the total weight coefficient of the target service:

[0185] The first proportion coefficient: It is related to the basic weight coefficient of service importance and reflects the proportion of service importance in the total weight. For example, if the importance of the service is crucial to the business, the first proportion coefficient will be larger, indicating that the weight of service importance is higher when calculating the total weight.

[0186] The second proportion coefficient: It is related to the basic weight coefficient of resource occupancy and reflects the proportion of resource occupancy in the total weight. If a certain service consumes a large amount of resources, the second proportion coefficient will be larger, indicating that the resource occupancy contributes more to the total weight.

[0187] Calculate the total weight coefficient: By the method of weighted average, combine the basic weight coefficient of service importance and the basic weight coefficient of resource occupancy. The specific formula is:

[0188] Total weight coefficient = (First proportion coefficient × Basic weight coefficient of service importance) + (Second proportion coefficient × Basic weight coefficient of resource occupancy)

[0189] This total weight coefficient will be used as the final weight of the target service in the category waiting area and determine the priority of the service in the resource allocation and optimization strategy.

[0190] Through the method provided in this embodiment, the total weight coefficient of the target service in the category waiting area can be determined scientifically and reasonably. First, the importance of the service is analyzed, and the basic weight coefficient of service importance is calculated; then, the resource occupancy situation is evaluated, and the basic weight coefficient of resource occupancy is obtained; finally, by combining the proportion coefficient, the total weight coefficient is calculated. This weight coefficient provides an important basis for subsequent resource allocation and optimization, ensuring that key services can be preferentially guaranteed and optimally resource-allocated.

[0191] In a second aspect, an adjustment system for a 5G network slice according to the present application includes:

[0192] A first acquisition unit 21, configured to acquire historical demand information of a target service;

[0193] A first determination unit 22, configured to determine an initial network slice strategy based on the historical demand information;

[0194] A second determination unit 23, configured to determine an initial network slice based on the initial network slice strategy;

[0195] A monitoring unit 24, configured to monitor performance data of each of the initial network slices;

[0196] A second acquisition unit 25, configured to acquire priority information of the target service corresponding to each network slice;

[0197] An adjustment unit 26, configured to make adjustments to the initial network slice based on the performance data and the priority information to determine an optimized network slice.

[0198] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for adjusting 5G network slices, characterized in that: include: Obtain historical demand information for target services; Determine an initial network slicing strategy based on the historical demand information; Based on the initial network slicing strategy, determine an initial network slice; Monitoring performance data of each of the initial network slices; Obtain the priority information of the target service corresponding to each network slice; Adjusting the initial network slice based on the performance data and the priority information to determine an optimized network slice; The determining of the initial network slicing strategy based on the historical demand information includes: A preset network slicing strategy is obtained by using a genetic algorithm based on the historical demand information, wherein the historical demand information includes bandwidth usage information, delay requirement information, and connection density information of a target service, the bandwidth usage information is obtained from a network log, the delay requirement information is obtained from a user usage record, and the connection density information is obtained from service load data; Determine an initial network slicing strategy according to the preset network slicing strategy through a reinforcement learning algorithm; The obtaining a preset network slicing strategy by using a genetic algorithm based on the historical demand information includes: Define an encoding method for the network slicing strategy, encode the bandwidth usage information, latency requirement information, and connection density information of each target service, and represent them as genes in the chromosome; Randomly extract typical resource allocation patterns from historical demand information and randomly generate an initial population; Design a fitness function to evaluate the quality of each chromosome; Performing selection operations, crossover operations, and mutation operations based on the fitness function, and continuously performing iterative operations until the fitness function improvement rate is less than a preset threshold or the number of iterations reaches a preset number, and outputting the preset network slicing strategy; The determining the initial network slicing strategy according to the preset network slicing strategy by using a reinforcement learning algorithm includes: The network environment is modeled as an environmental state in reinforcement learning, and the preset network slicing strategy is modeled as the behavior of an intelligent agent, wherein the environmental state includes the current network state, resource utilization, and service demand: Design a reward function to evaluate the quality of each action, where the reward function is: R(s,a)=δ·Q(s,a)-∈·D(s,a) Q(s,a) represents the service quality improvement obtained after taking action a in state s, D(s,a) represents the resource usage cost, and δ,∈ is the corresponding weight information; Adopt Q-learning reinforcement learning algorithm, aiming to maximize the cumulative reward, and update the strategy through repeated interaction with the environment; At each time step, the agent observes the current state, selects an action to adjust the network slicing strategy, and updates the network slicing strategy according to the reward function; When a preset number of iterations and / or strategy convergence is reached, outputting the initial network slicing strategy; The adjusting the initial network slice based on the performance data and the priority information to determine an optimized network slice includes: Defining a plurality of objective functions according to the performance data and the priority information, wherein the performance data includes delay data, bandwidth utilization data, energy consumption data, and service quality data; the plurality of objective functions include a delay objective function, a bandwidth utilization objective function, an energy consumption objective function, and a service quality objective function, each objective function includes a priority information, and the priority information is used to adjust the weight of the objective function; Use the NSGA-II algorithm to perform multi-objective optimization on all objective functions to adjust the initial network slice and determine the optimized network slice, wherein the optimized network slice includes a resource allocation slice, a quality of service slice, a bandwidth slice, a delay optimization slice, and an energy consumption optimization slice; The optimized network slice also includes a resource isolation slice and a resource protection slice; The method further comprises: Obtain performance bottleneck information and exception information; Determine a resource isolation slice and a resource protection slice based on the performance bottleneck information, the exception information and the priority information; The determining of the resource isolation slice and the resource protection slice based on the performance bottleneck information, the exception information and the priority information includes: Analyzing the performance bottleneck information and the abnormal information based on a hierarchical clustering algorithm to obtain multiple categories of candidate areas; Calculate the priority score of each category of the to-be-selected area according to the priority information corresponding to all target services in each category of the to-be-selected area; The resource isolation slice and resource protection slice are determined based on the priority score and priority score threshold of each category of candidate areas.

2. The adjustment method for 5G network slicing according to claim 1, characterized in that: The step of calculating the priority score of each category of the to-be-selected area according to the priority information corresponding to all target services in each category of the to-be-selected area comprises: The priority score S(Ci) is calculated according to the following formula: Among them, p ij Indicates the priority of the target service, w ij For C i The weight coefficient corresponding to the jth target service in the category to be selected area is, i is the index of the category to be selected area, and j is the index of a target service in the category to be selected area.

3. The adjustment method for 5G network slicing according to claim 2, characterized in that: The specific steps of determining the weight coefficients corresponding to the target services in each category of the selected area include: Determine the service importance basic weight coefficient corresponding to the target service in each category of the candidate area based on the number of service users, business revenue contribution and business continuity requirements of each target service in each category of the candidate area; Determine the resource occupancy basic weight coefficient corresponding to the target service in each category of the to-be-selected area based on the bandwidth information, CPU utilization information and storage space information of each target service in each category of the to-be-selected area; The weight coefficients corresponding to the target services in each category of the selected area are determined based on the service importance basic weight coefficient, the resource occupancy basic weight coefficient, the first weight coefficient and the second weight coefficient, wherein the first weight coefficient is the weight coefficient corresponding to the service importance basic weight coefficient, and the second weight coefficient is the weight coefficient corresponding to the resource occupancy basic weight coefficient.

4. A 5G network slice adjustment system, characterized in that: include: A first acquisition unit, used to acquire historical demand information of a target service; A first determining unit, configured to determine an initial network slicing strategy based on the historical demand information; A second determining unit, configured to determine an initial network slice based on the initial network slicing strategy; A monitoring unit, configured to monitor performance data of each of the initial network slices; A second acquisition unit is used to obtain priority information of a target service corresponding to each network slice; An adjusting unit, configured to adjust the initial network slice based on the performance data and the priority information to determine an optimized network slice; The determining of the initial network slicing strategy based on the historical demand information includes: A preset network slicing strategy is obtained by using a genetic algorithm based on the historical demand information, wherein the historical demand information includes bandwidth usage information, delay requirement information, and connection density information of a target service, the bandwidth usage information is obtained from a network log, the delay requirement information is obtained from a user usage record, and the connection density information is obtained from service load data; Determine an initial network slicing strategy according to the preset network slicing strategy through a reinforcement learning algorithm; The obtaining a preset network slicing strategy by using a genetic algorithm based on the historical demand information includes: Define an encoding method for the network slicing strategy, encode the bandwidth usage information, latency requirement information, and connection density information of each target service, and represent them as genes in the chromosome; Randomly extract typical resource allocation patterns from historical demand information and randomly generate an initial population; Design a fitness function to evaluate the quality of each chromosome; Performing selection operations, crossover operations, and mutation operations based on the fitness function, and continuously performing iterative operations until the fitness function improvement rate is less than a preset threshold or the number of iterations reaches a preset number, and outputting the preset network slicing strategy; The determining the initial network slicing strategy according to the preset network slicing strategy by using a reinforcement learning algorithm includes: The network environment is modeled as an environmental state in reinforcement learning, and the preset network slicing strategy is modeled as the behavior of an intelligent agent, wherein the environmental state includes the current network state, resource utilization, and service demand: Design a reward function to evaluate the quality of each action, where the reward function is: R(s,a)=δ·Q(s,a)-∈·D(s,a) Q(s,a) represents the service quality improvement obtained after taking action a in state s, D(s,a) represents the resource usage cost, and δ,∈ is the corresponding weight information; Adopt Q-learning reinforcement learning algorithm, aiming to maximize the cumulative reward, and update the strategy through repeated interaction with the environment; At each time step, the agent observes the current state, selects an action to adjust the network slicing strategy, and updates the network slicing strategy according to the reward function; When a preset number of iterations and / or strategy convergence is reached, outputting the initial network slicing strategy; The adjusting the initial network slice based on the performance data and the priority information to determine an optimized network slice includes: Defining a plurality of objective functions according to the performance data and the priority information, wherein the performance data includes delay data, bandwidth utilization data, energy consumption data, and service quality data; the plurality of objective functions include a delay objective function, a bandwidth utilization objective function, an energy consumption objective function, and a service quality objective function, each objective function includes a priority information, and the priority information is used to adjust the weight of the objective function; Use the NSGA-II algorithm to perform multi-objective optimization on all objective functions to adjust the initial network slice and determine the optimized network slice, wherein the optimized network slice includes a resource allocation slice, a quality of service slice, a bandwidth slice, a delay optimization slice, and an energy consumption optimization slice; The optimized network slice also includes a resource isolation slice and a resource protection slice; The adjustment unit also includes: Obtain performance bottleneck information and exception information; Determine a resource isolation slice and a resource protection slice based on the performance bottleneck information, the exception information and the priority information; The determining of the resource isolation slice and the resource protection slice based on the performance bottleneck information, the exception information and the priority information includes: Analyzing the performance bottleneck information and the abnormal information based on a hierarchical clustering algorithm to obtain multiple categories of candidate areas; Calculate the priority score of each category of the to-be-selected area according to the priority information corresponding to all target services in each category of the to-be-selected area; The resource isolation slice and resource protection slice are determined based on the priority score and priority score threshold of each category of candidate areas.

Images