Network slicing resource configuration method, device, equipment, storage medium and product

Abstract

The application provides a network slice resource configuration method, device, equipment, storage medium and product, generates a plurality of candidate configuration actions based on the network state parameters corresponding to the network slice collected at present; determines the SLA breach risk prediction value corresponding to each candidate configuration action; takes the candidate configuration action with the minimum SLA breach risk prediction value in the plurality of candidate configuration actions as the configuration action corresponding to the current network slice, and performs resource configuration on the network slice according to the configuration action corresponding to the current network slice. The scheme of the application can predict the SLA breach risk in the resource configuration process, and improves the security and reliability of the network.

Description

Technical Field

[0001] The present invention relates to the field of network configuration technology, and in particular to a network slicing resource configuration method, device, equipment, storage medium and product. Background Art

[0002] Network slicing uses software-defined networking and network function virtualization technologies to flexibly customize multiple virtual networks on the same physical network to meet the digital transmission services with differentiated service-level agreements (SLAs) requirements in vertical industries.

[0003] As a key business model for 5G and B5G, network slicing has begun to be applied to the Industrial Internet of Things (IIoT). Dynamically configuring network slice resources is key to improving network slicing performance. Current methods for slicing resource configuration fall into three main categories: optimization model-based, policy-based, and reinforcement learning-based.

[0004] In practice, scenarios with rapidly changing network environments, such as random user mobility, rapid channel degradation, and sudden service traffic, present a high risk of SLA breaches. Current slicing resource configuration methods are unaware of SLA breach risks, resulting in poor network security and reliability. Summary of the Invention

[0005] The present invention provides a network slice resource configuration method, device, equipment, storage medium and product to address the defect in the existing technology that the slice resource configuration method cannot perceive the SLA default risk, realize the SLA default risk prediction during the resource configuration process, and improve the security and reliability of the network.

[0006] The present invention provides a network slice resource configuration method, comprising: generating multiple candidate configuration actions based on network status parameters corresponding to the currently collected network slice; determining the SLA default risk prediction value corresponding to each candidate configuration action; taking the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions as the configuration action corresponding to the current network slice, and performing resource configuration on the network slice according to the configuration action corresponding to the current network slice.

[0007] According to a network slice resource configuration method provided by the present invention, the determining of the SLA default risk prediction value corresponding to each candidate configuration action includes: determining the slice resource configuration parameters corresponding to each candidate configuration action; inputting the slice resource configuration parameters corresponding to each candidate configuration action and the network status parameters corresponding to the currently collected network slice into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model; wherein the SLA default risk prediction model is constructed based on a machine learning algorithm and is a network model obtained by training based on a slice configuration experience data set; the slice configuration experience data set includes multiple historical slice configuration experience data; each historical slice configuration parameter includes historical network status parameters, historical slice resource configuration parameters and SLA default risk reference values ​​within the same slice time window.

[0008] According to a network slice resource configuration method provided by the present invention, a plurality of candidate configuration actions are generated based on the network status parameters corresponding to the currently collected network slice, including: generating a reference configuration action space based on the network status parameters corresponding to the currently collected network slice and a slice resource configuration experience database; wherein the slice resource configuration experience database includes a plurality of configuration experience data with an SLA default rate lower than a first threshold; and generating an initial configuration action based on the network status parameters corresponding to the currently collected network slice.

[0009] According to a network slice resource configuration method provided by the present invention, a reference configuration action space is generated based on the network status parameters corresponding to the currently collected network slice and the slice resource configuration experience database, including: extracting a reference configuration action from the slice resource configuration experience database according to the network status parameters corresponding to the currently collected network slice; and generating multiple extended configuration actions around the reference configuration action through an extended offset method to obtain the reference configuration action space.

[0010] According to a network slice resource configuration method provided by the present invention, the initial configuration action is generated based on the network status parameters corresponding to the currently acquired network slice, including: inputting the network status parameters corresponding to the currently acquired network slice into a slice resource configuration model, and obtaining the initial configuration action output by the slice resource configuration model; wherein, the slice resource configuration model is constructed based on a reinforcement learning algorithm.

[0011] According to a network slice resource configuration method provided by the present invention, after configuring resources for the network slice according to the configuration action corresponding to the current network slice, the method further includes: collecting slice performance parameters corresponding to the current network slice; wherein the slice performance parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, business service interruption rate, RB resource utilization, CPU computing resource utilization, and memory storage resource utilization; if the slice performance parameters do not meet the slice SLA requirements, then returning to the step of executing the network status parameters corresponding to the network slice obtained based on the current collection, and generating multiple configuration actions to be selected, until the slice performance parameters meet the slice SLA requirements.

[0012] The present invention also provides a network slice resource configuration device, including: a generation module, used to generate multiple candidate configuration actions based on the network status parameters corresponding to the currently collected network slice; a risk prediction module, used to determine the SLA default risk prediction value corresponding to each candidate configuration action; a configuration module, used to use the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions as the configuration action corresponding to the current network slice, and perform resource configuration on the network slice according to the configuration action corresponding to the current network slice.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the program, it implements any one of the network slice resource configuration methods described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements any of the network slice resource configuration methods described above.

[0015] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the network slice resource configuration methods described above.

[0016] The network slice resource configuration method, device, equipment, storage medium and product provided by the application determine the SLA violation risk prediction value corresponding to each candidate configuration action in the network slice resource configuration process, realize the SLA violation risk prediction of each candidate configuration action, further, take the candidate configuration action with the minimum SLA violation risk prediction value in the multiple candidate configuration actions as the configuration action corresponding to the current network slice, and perform resource configuration on the network slice according to the configuration action corresponding to the current network slice, which can reduce the SLA violation risk while ensuring the configuration requirement, therefore, the scheme of the application can perform SLA violation risk prediction in the resource configuration process, and improve the security and reliability of the network. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 is one of the flowcharts of the network slice resource configuration method provided by the application.

[0019] Figure 2 is the second flowchart of the network slice resource configuration method provided by the application.

[0020] Figure 3 is the third flowchart of the network slice resource configuration method provided by the application.

[0021] Figure 4 is the structural schematic diagram of the network slice resource configuration device provided by the application.

[0022] Figure 5 is the structural schematic diagram of the electronic equipment provided by the application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the application more clear, the technical scheme in the application will be described clearly and completely in the following by combining the drawings in the application. Obviously, the described embodiments are some embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0024] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes can not be described again in some examples. The following will be combined with Figure 1-Figure 3 The network slice resource configuration method of the present application is described.

[0025] Figure 1 is one of the flowcharts of the network slice resource configuration method provided by the present application, as shown in Figure 1 The method comprises the following steps 101, 102 and 103.

[0026] In practical applications, the execution subject of the network slice resource configuration method can be a network slice resource configuration device, and the implementation of the network slice resource configuration device has many ways, such as being realized by a computer program, for example, application software, etc.; or, for example, a chip, etc. It can also be realized as a medium with related computer programs stored, such as a U disk, a cloud disk, etc.; or, it can also be realized through an entity device integrated or installed with related computer programs, such as a server, etc.

[0027] Step 101, based on the network state parameters corresponding to the network slice collected at present, a plurality of candidate configuration actions are generated.

[0028] Among them, the network state parameter represents the network state of the network slice, as an example, the network state parameter comprises at least one of the following: uplink signal to interference plus noise ratio (SINR), channel quality indicator (CQI), modulation and coding strategy (MCS), uplink service traffic, user cache size.

[0029] As an example, the network state parameter comprises the service traffic value in the sliding slice window and the uplink SINR. In practical applications, the time length of each slice window can be dynamically set according to actual business needs, for example, the time length of the slice window is set to 1 second. The sliding slice window is composed of the previous several continuous slice windows of the current slice window, for example, 10 slice windows, and the time length of the corresponding sliding slice window can be set to 10 seconds.

[0030] The network slice resource configuration method provided by the embodiment can be applied to the resource configuration process of various service type network slices, such as an ultra-reliable and low latency (URLLC) service slice, a massive machine type of communication (mMTC) service slice, an enhanced mobile broadband (eMBB) service slice, and the like.

[0031] In the embodiment, the plurality of candidate configuration actions can be generated in different ways according to the network state parameters of the current network slice collected. The plurality of candidate configuration actions have good adaptability to the current network state. As an example, a slice resource configuration experience database can be pre-set according to historical configuration experience. Specifically, the slice resource configuration experience database is queried based on the network state parameters to obtain the candidate configuration action. At the same time, the reinforcement learning method is used to determine the candidate configuration action based on the network state parameters.

[0032] Step 102, determining a SLA breach risk prediction value corresponding to each candidate configuration action.

[0033] The SLA breach risk prediction value refers to the prediction value of the possibility of the SLA breach. In actual application, a slice performance statistical value representing the SLA breach can be a slice SLA breach rate. As an example, the SLA breach risk prediction value can be the prediction value of the slice SLA breach rate. The calculation method of the slice SLA breach rate is 1 minus the product of the slice SLA satisfaction rate of the current slice window. The slice SLA satisfaction rate refers to the ratio of the number of data packets that simultaneously meet the delay and user rate requirements in the data packets generated by the service in the current slice window to the total number of all data packets generated in the current slice window, multiplied by the return value of the packet loss rate determination function in the slice window. In the packet loss rate determination function, if the packet loss rate is not higher than the packet loss rate threshold, the return result is 1, otherwise the return result is 0. The value range of the above-mentioned slice SLA satisfaction rate and slice SLA breach rate is [0, 1].

[0034] It can be understood that the SLA breach risk prediction value corresponding to each candidate configuration action is determined respectively, and the possibility of SLA breach of each candidate configuration action can be determined. Further, the configuration action corresponding to the current network slice can be determined from the plurality of candidate configuration actions according to the SLA breach risk prediction value corresponding to each candidate configuration action.

[0035] Step 103: The candidate configuration action with the smallest SLA default risk prediction value among multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and resources are configured for the network slice according to the configuration action corresponding to the current network slice.

[0036] In this embodiment, the configuration action with the lowest predicted SLA default risk among multiple candidate configuration actions is selected as the configuration action corresponding to the current network slice. This allows the configuration action with the lowest SLA default risk to be selected while ensuring the resource configuration requirements of the network slice. Furthermore, resource configuration for the network slice based on the configuration action corresponding to the current network slice can effectively improve network security and reliability.

[0037] In addition, regarding the method for determining the SLA default risk prediction value, as an example, in a possible implementation, Figure 2 This is the second flow chart of the network slice resource configuration method provided by the present invention, as shown in FIG. Figure 2 As shown, the above step 102 includes: step 201 and step 202.

[0038] Step 201: Determine the slice resource configuration parameters corresponding to each candidate configuration action.

[0039] Slice resource configuration parameters refer to parameters used for network configuration. As an example, slice resource configuration parameters can be a combination of network resources reserved for different slices within the current slice window, including the number of physical resource blocks (RBs), the number of central processing units (CPUs), and the number of memory resources.

[0040] In another example, given the total amount of RB, CPU, and memory resources, it is necessary to dynamically configure the ratio of RB, CPU, and memory resources within each slice to reduce the risk of slice SLA breach, improve the stability of slice satisfaction rate, and reduce network performance jitter. Specifically, the slice resource configuration parameters include the ratio of the number of RBs configured in each slice to the total number of uplink RBs, the ratio of the number of CPU cycles configured in each slice to the total number of CPU cycles, and the ratio of the number of memory configured in each slice to the total number of memories.

[0041] Step 202: Input the slice resource configuration parameters corresponding to each candidate configuration action and the network status parameters corresponding to the currently collected network slice into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model.

[0042] The SLA breach risk prediction model is constructed based on a machine learning algorithm and is a network model trained based on a slice configuration experience data set. The slice configuration experience data set includes a plurality of historical slice configuration experience data. Each piece of historical slice configuration parameter includes historical network state parameters, historical slice resource configuration parameters, and an SLA breach risk reference value in a same slice time window.

[0043] In actual application, the slice configuration experience data set can be constructed based on data collected in a historical period. In an example, the network slice resource configuration method further includes: constructing the slice configuration experience data set based on the collected historical network state parameters, historical slice resource configuration parameters, and historical slice performance parameters.

[0044] Specifically, the collected historical network state parameters, historical slice resource configuration parameters, and historical slice performance parameters are centrally processed, for example, parameter normalization and standardization processing. Further, based on the historical slice performance parameters in each slice time window, the SLA breach reference value in the slice time window is determined; the historical network state parameters, historical slice resource configuration parameters, and SLA breach reference value in each slice time window are taken as a piece of historical slice configuration parameter, and the configuration experience data set is obtained.

[0045] In actual application, an initial SLA breach risk prediction model is constructed based on a machine learning algorithm, and the initial SLA breach risk prediction model is trained according to the slice configuration experience data set. In this embodiment, the machine learning algorithm applied to construct the initial SLA breach risk prediction model is not specifically limited, and as an example, the initial SLA breach risk prediction model can be established based on a deep neural network (Deep Neural Networks, DNN for short), a recurrent neural network (Recurrent Neural Network, RNN for short), a long short-term memory unit (Long Short-Term Memory, LSTM for short), a gated recurrent unit (Gated Recurrent Unit, GRU for short), and the like.

[0046] The following briefly describes the process of training an SLA default risk prediction model, using the DNN algorithm as an example. An initial SLA default risk prediction model is constructed based on the DNN algorithm. The network structure of the initial SLA default risk prediction model consists of an input layer, N hidden layers, and an output layer. During training, the input to the current SLA default risk prediction model is historical network state parameters and historical slice resource configuration parameters. The output of the current SLA default risk prediction model is the predicted SLA default risk value. The input layer of the SLA default risk prediction model is connected to the first hidden layer, and the neurons in the first hidden layer are connected to the neurons in the next hidden layer. Similarly, the neurons in the Nth hidden layer are connected to the output layer, which ultimately outputs the predicted SLA default risk value.

[0047] During the training process of the SLA default risk prediction model, the prediction error is back-propagated through the supervised learning mechanism, and the weight and bias parameters of the current SLA default risk prediction model are gradually adjusted to continuously reduce the difference between the SLA default risk prediction value and the SLA default risk reference value until the error meets the actual needs and the SLA default risk prediction model is obtained.

[0048] It is understandable that after the training is completed to obtain the SLA default risk prediction model, the SLA default risk prediction value corresponding to each candidate configuration action can be predicted by the SLA default risk prediction model. As an example, the slice resource configuration parameters include the ratio of the number of RB configurations of each slice to the total number of uplink RBs, the ratio of the periodic configuration amount of each slice CPU to the total number of CPU periods, and the ratio of the memory configuration amount of each slice to the total number of memories. The network state parameters corresponding to the currently collected network slice include the service flow and uplink SINR of the sliding slice window at the current moment. The SLA default risk prediction value is the predicted value of the SLA default rate. Specifically, step 202 includes: inputting the service flow of the sliding slice window at the current moment, the uplink SINR, the ratio of the number of RB configurations of each slice to the total number of uplink RBs, the ratio of the periodic configuration amount of each slice CPU to the total number of CPU periods, and the ratio of the memory configuration amount of each slice to the total number of memories into the SLA default risk prediction model to obtain the predicted value of the SLA default rate corresponding to each candidate configuration action output by the SLA default risk prediction model.

[0049] In this embodiment, the slice resource configuration parameters corresponding to each candidate configuration action and the network state parameters corresponding to the currently collected network slice are input into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model, which can predict the slice SLA default risk of different candidate configuration actions under different network states. On this basis, the candidate configuration action with the smallest SLA default risk prediction value among multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and the network slice is resource configured according to the configuration action corresponding to the current network slice, which can reduce the SLA default risk while ensuring the configuration requirements; therefore, the solution of this embodiment can predict the SLA default risk during the resource configuration process, thereby improving the security and reliability of the network.

[0050] In addition, as an example of a method for generating multiple candidate configuration actions, in a possible implementation, Figure 3 This is the third flow chart of the network slice resource configuration method provided by the present invention, as shown in FIG. Figure 3 As shown, the above step 101 includes: step 301 and step 302.

[0051] Step 301: Generate a reference configuration action space based on the network status parameters corresponding to the currently collected network slice and the slice resource configuration experience database.

[0052] The slice resource configuration experience database includes multiple configuration experience data sets with SLA default rates below a first threshold. In practical applications, an SLA default rate below the first threshold indicates that the SLA default risk of historical configuration actions is low. For example, the first threshold can be set to 0.02.

[0053] Combined with the above description, the slice SLA default rate is calculated by subtracting the product of the SLA satisfaction rates of each slice in the current slice window from the slice SLA default rate in method 1. In one example, the slice resource configuration experience database includes configuration experience data of multiple slices with high SLA satisfaction rates, for example, configuration experience data of slices with SLA satisfaction rates higher than 0.98.

[0054] In practical applications, a slice resource configuration experience database can be constructed based on data collected over historical periods. In one example, the network slice resource configuration method further includes: constructing a slice resource configuration experience database based on collected historical network status parameters, historical slice resource configuration parameters, and historical slice performance parameters.

[0055] Specifically, the obtained historical network state parameters, historical slice resource configuration parameters, and historical slice performance parameters are collected. It can be understood that the historical period includes a plurality of historical collection data, and each historical collection data includes historical network state parameters, historical slice resource configuration parameters, and historical slice performance parameters. Specifically, for each historical collection data, the SLA violation reference value is determined according to the historical slice performance parameters, and the historical configuration action is determined according to the historical slice resource configuration parameters. Further, the historical collection data with the SLA violation reference value lower than the first threshold value is selected to construct the slice resource configuration experience database. It can be understood that each configuration experience data in the slice resource configuration experience database includes: historical network state parameters, historical slice resource configuration parameters, historical slice performance parameters, SLA violation reference value, and historical configuration action.

[0056] As an example, the above step 301 includes: according to the network state parameters corresponding to the current collected network slice, extracting the reference configuration action from the slice resource configuration experience database. By extending the offset method, a plurality of extended configuration actions around the reference configuration action are generated to obtain the reference configuration action space.

[0057] In practical applications, according to the network state parameters corresponding to the current collected network slice, the corresponding configuration experience data is extracted from the slice resource configuration experience database by the network priori knowledge analysis method, such as the nearest neighbor search method, and the reference configuration action is output.

[0058] In one example, the Euclidean distance between the network state parameters corresponding to the current collected network slice and the historical network state parameters in each configuration experience data in the slice resource configuration experience database is calculated, the configuration experience data with the smallest Euclidean distance is obtained, the historical configuration action in the configuration experience data is extracted, and the historical configuration action is taken as the reference configuration action.

[0059] Further, considering the time-varying nature of the network environment, the offset range of the slice resource configuration parameters corresponding to the reference configuration action is expanded by the extended offset method to obtain a plurality of extended configuration actions around the reference configuration action, and finally the slice reference configuration action space is generated.

[0060] In one example, the maximum offset ratio of each dimension of the reference configuration action on each slice is selected by using the extended offset method. The dimensions of the reference configuration action are respectively the proportion of the number of RB configurations on each slice to the total number of uplink RBs, the proportion of the periodic configuration amount of CPU to the total amount of CPU, and the proportion of the memory configuration amount to the total amount of memory. Specifically, the maximum offset of the slice resource configuration ratio on each dimension corresponding to the maximum offset can be set to ±15% by using the extended offset method, and all possible offsets [-15%, -10%, -5%, 0, 5%, 10%, 15%] are traversed with a step of 5%, to obtain a plurality of extended configuration actions around the reference configuration action, and finally the slice reference configuration action space composed of all the extended configuration actions is output.

[0061] In this example, considering the influence of effective configuration experience of network environment changes, the slice resource configuration facing low SLA violation risk provides the slice reference configuration action space, thereby appropriately expanding the scanning range of the candidate configuration action, providing a reference for selecting the slice configuration action with the lowest SLA violation risk, and being conducive to realizing the low-risk slice resource configuration decision.

[0062] Step 302, generating an initial configuration action based on the network state parameters corresponding to the network slice currently collected.

[0063] As an example, the above step 302 includes: inputting the network state parameters corresponding to the network slice currently collected into the slice resource configuration model to obtain the initial configuration action output by the slice resource configuration model; wherein the slice resource configuration model is constructed based on a reinforcement learning algorithm.

[0064] In this embodiment, the reinforcement learning algorithm applied to construct the slice resource configuration model is not specifically limited, and as an example, the initial slice resource configuration model can be established based on a reinforcement learning network such as Deep Q-Network (DQN), Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), etc.

[0065] The following uses the DQN algorithm as an example to briefly describe the process of outputting the initial configuration action through the slice resource configuration model. Specifically, in the DQN-based slice resource configuration model algorithm, the service traffic of each slice in the current sliding slice window and the uplink SINR serve as the input of the slice resource configuration model. DQN uses the ε-Greedy strategy to select the original configuration action. This strategy uses a small probability ε to select a random slice configuration action from the complete slice configuration action space (all permutations and combinations of the RB, CPU, and memory configuration ratios of each slice under the total resource limit). The slice configuration action with the highest slice SLA satisfaction rate is selected with a probability of 1-ε to obtain the initial configuration action. ε can be set to 0.02.

[0066] In practical applications, after resource configuration is completed, the slice performance parameters corresponding to the current network slice can be collected and fed back to the slice resource configuration model as a network reward for the slice resource configuration model. The slice performance parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, service interruption rate, RB resource utilization, CPU computing resource utilization, and memory storage resource utilization.

[0067] In this implementation, a reference configuration action space is generated based on the network status parameters corresponding to the currently acquired network slice and the slice resource configuration experience database; an initial configuration action is generated based on the network status parameters corresponding to the currently acquired network slice, and a plurality of candidate configuration actions are obtained; on this basis, the SLA default risk prediction value corresponding to each candidate configuration action is determined, thereby realizing the SLA default risk prediction for each candidate configuration action; the candidate configuration action with the smallest SLA default risk prediction value among the plurality of candidate configuration actions is used as the configuration action corresponding to the current network slice, and resources are configured for the network slice according to the configuration action corresponding to the current network slice, thereby reducing the SLA default risk while ensuring the configuration requirements; the SLA default risk prediction can be performed during the resource configuration process, thereby improving the security and reliability of the network.

[0068] Furthermore, in order to further improve the security and reliability of the network, after configuring the resources for the network slice, it is possible to predict whether the current network status meets the requirements. As an example, in one possible implementation, after the above step 103, the network slice resource configuration method further includes: collecting slice performance parameters corresponding to the current network slice; wherein the slice performance parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, business service interruption rate, RB resource utilization, CPU computing resource utilization, and memory storage resource utilization; if the slice performance parameters do not meet the slice SLA requirements, then return to the step of generating multiple candidate configuration actions based on the network status parameters corresponding to the currently collected network slice until the slice performance parameters meet the slice SLA requirements.

[0069] In practical applications, the slice SLA requirement parameters corresponding to the current network slice are collected and then the slice SLA requirements for the current network slice are determined based on the slice SLA requirement parameters. The slice SLA requirement parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, and service interruption rate. Furthermore, a determination is made as to whether the slice performance statistics meet the slice SLA requirements. If not, resource allocation for the network slice continues.

[0070] Optionally, after collecting and obtaining the slice performance parameters corresponding to the current network slice, slice performance statistics can be performed based on the slice performance parameters corresponding to the current network slice to obtain slice performance statistical parameters. As an example, a statistical method based on a slice time window can be adopted, with an independent slice time window as the time range for slice performance parameter statistics, for example, the time range is set to 1 second. The slice performance statistical parameters include at least one of the following: throughput, latency, bit error rate, packet loss rate, slice SLA satisfaction rate, slice SLA default rate, and business service interruption rate. Further, it is determined whether the slice performance statistical parameters meet the slice SLA requirements. If not, continue to configure resources for the network slice.

[0071] In the network slice resource configuration method provided by this embodiment, based on the network state parameters corresponding to the currently collected network slice, multiple candidate configuration actions are generated; the SLA default risk prediction value corresponding to each candidate configuration action is determined; the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and the network slice is configured with resources according to the configuration action corresponding to the current network slice. The solution of this embodiment determines the SLA default risk prediction value corresponding to each candidate configuration action during the network slice resource configuration process, thereby realizing the SLA default risk prediction for each candidate configuration action; further, the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and the network slice is configured with resources according to the configuration action corresponding to the current network slice, which can reduce the SLA default risk while ensuring the configuration requirements; therefore, the solution of this embodiment can perform SLA default risk prediction during the resource configuration process, thereby improving the security and reliability of the network.

[0072] The network slice resource configuration device provided by the present invention is described below. The network slice resource configuration device described below and the network slice resource configuration method described above can be referenced to each other.

[0073] Figure 4 This is a schematic diagram of the structure of the network slice resource configuration device provided by the present invention. Figure 4 As shown, the network slice resource configuration device includes: a generation module 41, a risk prediction module 42 and a configuration module 43.

[0074] The above-mentioned generation module 41 is used to generate multiple candidate configuration actions based on the network status parameters corresponding to the currently collected network slice.

[0075] Among them, the network status parameters characterize the network status of the network slice. As an example, the network status parameters include at least one of the following: uplink signal-to-interference and noise ratio (SINR), channel quality indication (CQI), modulation and coding strategy (MCS), uplink service traffic, and user buffer size.

[0076] In this embodiment, the multiple candidate configuration actions can be generated by the generation module 41 in different ways based on the network state parameters corresponding to the currently acquired network slice. The multiple candidate configuration actions are all well adapted to the current network state.

[0077] The risk prediction module 42 is configured to determine an SLA breach risk prediction value corresponding to each candidate configuration action.

[0078] The SLA default risk prediction value refers to a predicted value of the likelihood of an SLA default. In practical applications, a slice performance statistic that characterizes an SLA default situation may be a slice SLA default rate. As an example, the SLA default risk prediction value may be a predicted value of a slice SLA default rate.

[0079] It is understood that by determining the corresponding SLA breach risk prediction value for each candidate configuration action, the possibility of SLA breach for each candidate configuration action can be determined. Furthermore, based on the SLA breach risk prediction value corresponding to each candidate configuration action, the configuration action corresponding to the current network slice can be determined from multiple candidate configuration actions.

[0080] The configuration module 43 is used to select the candidate configuration action with the smallest SLA default risk prediction value among multiple candidate configuration actions as the configuration action corresponding to the current network slice, and perform resource configuration on the network slice according to the configuration action corresponding to the current network slice.

[0081] In this embodiment, the configuration module 43 selects the configuration action with the lowest predicted SLA default risk value among multiple candidate configuration actions as the configuration action corresponding to the current network slice. This allows the configuration action with the lowest SLA default risk to be selected while ensuring the resource configuration requirements of the network slice. Furthermore, the configuration module 43 configures resources for the network slice based on the configuration action corresponding to the current network slice, effectively improving network security and reliability.

[0082] As an example, in one possible implementation, the risk prediction module 42 is specifically configured to determine the slice resource configuration parameters corresponding to each candidate configuration action. The slice resource configuration parameters corresponding to each candidate configuration action and the network status parameters corresponding to the currently acquired network slice are input into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model.

[0083] Slice resource configuration parameters refer to parameters used for network configuration. As an example, the slice resource configuration parameters can be a combination of network resources reserved for different slices within the current slice window, including the number of physical resource blocks (RBs), the number of computing processing units (CPUs), and the number of memory resources.

[0084] In another example, given the total amount of RB, CPU, and memory resources, it is necessary to dynamically configure the ratio of RB, CPU, and memory resources within each slice to reduce the risk of slice SLA breach, improve the stability of slice satisfaction rate, and reduce network performance jitter. Specifically, the slice resource configuration parameters include the ratio of the number of RBs configured in each slice to the total number of uplink RBs, the ratio of the number of CPU cycles configured in each slice to the total number of CPU cycles, and the ratio of the number of memory configured in each slice to the total number of memories.

[0085] Among them, the SLA default risk prediction model is built based on a machine learning algorithm and is a network model trained based on a slice configuration experience dataset; the slice configuration experience dataset includes multiple historical slice configuration experience data; each historical slice configuration parameter includes the historical network status parameters, historical slice resource configuration parameters and SLA default risk reference value within the same slice time window.

[0086] In practical applications, the slice configuration experience dataset can be constructed based on data collected over a historical period. In one example, the network slice resource configuration device further includes: a first construction module for constructing a slice configuration experience dataset based on the collected historical network status parameters, historical slice resource configuration parameters, and historical slice performance parameters.

[0087] Specifically, the collected historical network status parameters, historical slice resource configuration parameters, and historical slice performance parameters are centrally processed, for example, by parameter normalization and standardization. Furthermore, based on the historical slice performance parameters within each slice time window, the SLA breach reference value within that slice time window is determined. The historical network status parameters, historical slice resource configuration parameters, and SLA breach reference value within each slice time window are used as a historical slice configuration parameter to obtain a configuration experience dataset.

[0088] In practical applications, an initial SLA default risk prediction model is constructed based on a machine learning algorithm, and the initial SLA default risk prediction model is trained based on a slice configuration experience dataset. In this embodiment, the machine learning algorithm used to construct the initial SLA default risk prediction model is not specifically limited.

[0089] During the training process of the SLA default risk prediction model, the prediction error is back-propagated through the supervised learning mechanism, and the weight and bias parameters of the current SLA default risk prediction model are gradually adjusted to continuously reduce the difference between the SLA default risk prediction value and the SLA default risk reference value until the error meets the actual needs and the SLA default risk prediction model is obtained.

[0090] In this embodiment, the slice resource configuration parameters corresponding to each candidate configuration action and the network state parameters corresponding to the currently collected network slice are input into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model, which can predict the slice SLA default risk of different candidate configuration actions under different network states. On this basis, the candidate configuration action with the smallest SLA default risk prediction value among multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and the network slice is resource configured according to the configuration action corresponding to the current network slice, which can reduce the SLA default risk while ensuring the configuration requirements; therefore, the solution of this embodiment can predict the SLA default risk during the resource configuration process, thereby improving the security and reliability of the network.

[0091] As an example, in a possible implementation, the generation module 41 includes: a first generation unit and a second generation unit.

[0092] The first generation unit is used to generate a reference configuration action space based on the network status parameters corresponding to the currently collected network slice and the slice resource configuration experience database.

[0093] The slice resource configuration experience database includes multiple configuration experience data with SLA default rates lower than a first threshold. In practical applications, an SLA default rate lower than the first threshold indicates that the SLA default risk of historical configuration actions is low.

[0094] In practical applications, the slice resource configuration experience database can be constructed based on data collected during historical periods. In one example, the network slice resource configuration device further includes: a second construction module for constructing a slice resource configuration experience database based on the collected historical network status parameters, historical slice resource configuration parameters, and historical slice performance parameters.

[0095] Specifically, the collected historical network status parameters, historical slice resource configuration parameters and historical slice performance parameters are obtained. It is understandable that the historical period includes multiple historical collection data, and each historical collection data includes historical network status parameters, historical slice resource configuration parameters and historical slice performance parameters. Specifically, for each historical collection data, the SLA breach reference value is determined based on the historical slice performance parameters; the historical configuration action is determined based on the historical slice resource configuration parameters. Furthermore, historical collection data with an SLA breach reference value lower than a first threshold is selected to construct a slice resource configuration experience database. It is understandable that each configuration experience data in the slice resource configuration experience database includes: historical network status parameters, historical slice resource configuration parameters, historical slice performance parameters, SLA breach reference value and historical configuration action.

[0096] As an example, the first generation unit is specifically configured to: extract a reference configuration action from a slice resource configuration experience database based on the network state parameters corresponding to the currently acquired network slice. By using an extended offset method, multiple extended configuration actions surrounding the reference configuration action are generated to obtain a reference configuration action space.

[0097] In practical applications, based on the network status parameters corresponding to the currently collected network slice, the corresponding configuration experience data is extracted from the slice resource configuration experience database through network prior knowledge analysis methods, such as the nearest neighbor search method, and the reference configuration action is output.

[0098] In an example, the Euclidean distance between the network status parameters corresponding to the currently collected network slice and the historical network status parameters in each configuration experience data in the slice resource configuration experience database is calculated to obtain the configuration experience data with the smallest Euclidean distance, extract the historical configuration action in the configuration experience data, and use the historical configuration action as a reference configuration action.

[0099] Furthermore, considering the time-varying nature of the network environment, an extended offset method is adopted to expand the offset range of the slice resource configuration parameters corresponding to the reference configuration action, obtain multiple extended configuration actions around the reference configuration action, and finally generate the slice reference configuration action space.

[0100] In one example, an extended offset method is used to select the maximum offset ratio of each dimension on each slice of the reference configuration action. The dimensions of the reference configuration action are the ratio of the number of RB configurations on each slice to the total number of uplink RBs, the ratio of the CPU cycle configuration to the total number of CPU cycles, and the ratio of the memory configuration to the total amount of memory. Specifically, the maximum offset corresponding to the slice resource configuration ratio on each dimension can be set by the extended offset method to obtain multiple extended configuration actions around the reference configuration action, and finally output the slice reference configuration action space composed of all extended configuration actions.

[0101] In this example, the impact of effective configuration experience due to changes in the network environment is considered. A slice reference configuration action space is provided for slice resource configuration with low SLA breach risk, thereby appropriately expanding the scanning range of candidate configuration actions. This provides a reference for selecting the slice configuration action with the lowest SLA breach risk, which is conducive to achieving low-risk slice resource configuration decisions.

[0102] The above-mentioned second generation unit is used to generate an initial configuration action based on the network status parameters corresponding to the currently collected network slice.

[0103] As an example, the above-mentioned second generation unit is specifically used to: input the network status parameters corresponding to the currently collected network slice into the slice resource configuration model to obtain the initial configuration action output by the slice resource configuration model; wherein, the slice resource configuration model is constructed based on the reinforcement learning algorithm.

[0104] In this embodiment, there is no specific limitation on the reinforcement learning algorithm used to build the slice resource configuration model.

[0105] In actual applications, after completing resource configuration, the slice performance parameters corresponding to the current network slice can be collected and fed back to the slice resource configuration model as a network reward for the slice resource configuration model.

[0106] In this implementation, a reference configuration action space is generated based on the network status parameters corresponding to the currently acquired network slice and the slice resource configuration experience database; an initial configuration action is generated based on the network status parameters corresponding to the currently acquired network slice, and a plurality of candidate configuration actions are obtained; on this basis, the SLA default risk prediction value corresponding to each candidate configuration action is determined, thereby realizing the SLA default risk prediction for each candidate configuration action; the candidate configuration action with the smallest SLA default risk prediction value among the plurality of candidate configuration actions is used as the configuration action corresponding to the current network slice, and resources are configured for the network slice according to the configuration action corresponding to the current network slice, thereby reducing the SLA default risk while ensuring the configuration requirements; the SLA default risk prediction can be performed during the resource configuration process, thereby improving the security and reliability of the network.

[0107] As an example, in one possible implementation, the network slice resource configuration device further includes: a detection module for collecting slice performance parameters corresponding to the current network slice; wherein the slice performance parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, business service interruption rate, RB resource utilization, CPU computing resource utilization, and memory storage resource utilization; if the slice performance parameters do not meet the slice SLA requirements, then the step of returning to execute the network status parameters corresponding to the currently collected network slice to generate multiple configuration actions until the slice performance parameters meet the slice SLA requirements.

[0108] In practical applications, the slice SLA requirement parameters corresponding to the current network slice are collected and then the slice SLA requirements for the current network slice are determined based on the slice SLA requirement parameters. The slice SLA requirement parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, and service interruption rate. Furthermore, a determination is made as to whether the slice performance statistics meet the slice SLA requirements. If not, resource allocation for the network slice continues.

[0109] Optionally, after collecting and obtaining the slice performance parameters corresponding to the current network slice, slice performance statistics can be performed based on the slice performance parameters corresponding to the current network slice to obtain slice performance statistical parameters. As an example, a statistical method based on a slice time window can be adopted, with an independent slice time window as the time range for slice performance parameter statistics. Slice performance statistical parameters include at least one of the following: throughput, latency, bit error rate, packet loss rate, slice SLA satisfaction rate, slice SLA default rate, and business service interruption rate. Further, it is determined whether the slice performance statistical parameters meet the slice SLA requirements. If not, continue to configure resources for the network slice.

[0110] In the network slice resource configuration device provided by this embodiment, the generation module generates multiple candidate configuration actions based on the network state parameters corresponding to the currently acquired network slice; the risk prediction module determines the SLA default risk prediction value corresponding to each candidate configuration action; the configuration module uses the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions as the configuration action corresponding to the current network slice, and performs resource configuration for the network slice according to the configuration action corresponding to the current network slice. The solution of this embodiment determines the SLA default risk prediction value corresponding to each candidate configuration action during the network slice resource configuration process, thereby realizing the SLA default risk prediction for each candidate configuration action; further, the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and performs resource configuration for the network slice according to the configuration action corresponding to the current network slice, which can reduce the SLA default risk while ensuring the configuration requirements; therefore, the solution of this embodiment can perform SLA default risk prediction during the resource configuration process, thereby improving the security and reliability of the network.

[0111] Figure 5 An example of a physical structure diagram of an electronic device is shown below. Figure 5As shown, the electronic device may include: a processor (processor) 510, a communication interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 530 to execute the network slice resource configuration method, which includes: based on the network state parameters corresponding to the currently acquired network slice, generating multiple candidate configuration actions. Determine the SLA default risk prediction value corresponding to each candidate configuration action. The candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and according to the configuration action corresponding to the current network slice, the network slice is configured with resources.

[0112] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0113] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the network slice resource configuration method provided by the above methods, and the method includes: based on the network status parameters corresponding to the currently acquired network slice, generating multiple candidate configuration actions. Determine the SLA default risk prediction value corresponding to each candidate configuration action. The candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and according to the configuration action corresponding to the current network slice, the network slice is configured with resources.

[0114] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the network slice resource configuration method provided by each of the above methods, and the method comprises: generating a plurality of candidate configuration actions based on the network state parameters corresponding to the network slice currently collected; determining a SLA breach risk prediction value corresponding to each candidate configuration action; taking the candidate configuration action with the minimum SLA breach risk prediction value among the plurality of candidate configuration actions as the configuration action corresponding to the current network slice, and performing resource configuration on the network slice according to the configuration action corresponding to the current network slice.

[0115] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0116] From the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in terms of contribution to the prior art, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0117] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A network slice resource configuration method, characterized in that: include: Generate multiple candidate configuration actions based on the network status parameters corresponding to the currently collected network slice; Determine the SLA breach risk prediction value corresponding to each candidate configuration action; The candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions is used as the configuration action corresponding to the current network slice, and resources are configured for the network slice according to the configuration action corresponding to the current network slice; Determining the SLA default risk prediction value corresponding to each candidate configuration action includes: Determine the slice resource configuration parameters corresponding to each candidate configuration action; The slice resource configuration parameters corresponding to each candidate configuration action and the network status parameters corresponding to the currently collected network slice are input into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model.

2. The network slice resource configuration method according to claim 1, characterized in that: The SLA default risk prediction model is constructed based on a machine learning algorithm and is a network model obtained by training based on a slice configuration experience dataset; the slice configuration experience dataset includes multiple historical slice configuration experience data; Each historical slice configuration parameter includes historical network status parameters, historical slice resource configuration parameters, and SLA breach risk reference values ​​within the same slice time window.

3. The network slice resource configuration method according to claim 1, characterized in that: The generating of multiple configuration actions based on the network state parameters corresponding to the currently acquired network slice includes: Generate a reference configuration action space based on the network state parameters corresponding to the currently acquired network slice and a slice resource configuration experience database; wherein the slice resource configuration experience database includes a plurality of configuration experience data with an SLA default rate lower than a first threshold; Based on the network status parameters corresponding to the currently collected network slice, an initial configuration action is generated.

4. The network slice resource configuration method according to claim 3, characterized in that: The generating a reference configuration action space based on the network state parameters corresponding to the currently acquired network slice and the slice resource configuration experience database includes: Extracting a reference configuration action from a slice resource configuration experience database based on the network status parameters corresponding to the currently acquired network slice; A plurality of extended configuration actions surrounding the reference configuration action are generated by an extended offset method to obtain the reference configuration action space.

5. The network slice resource configuration method according to claim 3, characterized in that: The generating an initial configuration action based on the network state parameter corresponding to the currently acquired network slice includes: The network status parameters corresponding to the currently collected network slice are input into a slice resource configuration model to obtain an initial configuration action output by the slice resource configuration model; wherein the slice resource configuration model is constructed based on a reinforcement learning algorithm.

6. The network slice resource configuration method according to any one of claims 1 to 5, characterized in that: After configuring resources for the network slice according to the configuration action corresponding to the current network slice, the method further includes: Collect slice performance parameters corresponding to the current network slice; wherein the slice performance parameters include at least one of the following: user rate, latency, bit error rate, packet loss rate, service interruption rate, RB resource utilization, CPU computing resource utilization, and memory storage resource utilization; If the slice performance parameters do not meet the slice SLA requirements, return to the step of executing the network status parameters corresponding to the network slice currently collected, and generate multiple optional configuration actions until the slice performance parameters meet the slice SLA requirements.

7. A network slice resource configuration device, characterized in that: include: A generation module is used to generate multiple candidate configuration actions based on the network status parameters corresponding to the currently acquired network slice; A risk prediction module is used to determine the SLA breach risk prediction value corresponding to each candidate configuration action; a configuration module, configured to select the candidate configuration action with the smallest SLA default risk prediction value among the multiple candidate configuration actions as the configuration action corresponding to the current network slice, and perform resource configuration on the network slice according to the configuration action corresponding to the current network slice; The risk prediction module is specifically used to: Determine the slice resource configuration parameters corresponding to each candidate configuration action; The slice resource configuration parameters corresponding to each candidate configuration action and the network status parameters corresponding to the currently collected network slice are input into the SLA default risk prediction model to obtain the SLA default risk prediction value corresponding to each candidate configuration action output by the SLA default risk prediction model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the network slice resource configuration method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the network slice resource configuration method as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the network slice resource configuration method as described in any one of claims 1 to 6.

Images