Network switch control method and system

Abstract

The invention discloses a network switch control method and system, particularly relates to the technical field of network flow management, and is used for solving the problem that low-frequency flow is suddenly lost or rearranged when high-frequency flow is dynamically compressed in the prior art. The method comprises the following steps: monitoring all traffic entering a network switch to obtain frequency characteristics of the traffic, and classifying the traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics; compressing the high-frequency traffic by applying a dynamic traffic compression algorithm; for the low-frequency flow, analyzing a topological link by adopting a sparse representation iteration reweighted least square method, and identifying the interference intensity of the link; evaluating the logical coherence of the low-frequency flow data packet through a variational auto-encoder; determining a priority scheduling strategy of the low-frequency traffic based on the link interference intensity and the logic coherence; and finally, carrying out resource allocation and scheduling on the compressed high-frequency flow and the preferentially processed low-frequency flow.

Description

Technical Field

[0001] The present invention relates to the technical field of network traffic management, and more specifically, to a network switch control method and system. Background Art

[0002] In modern network communications, network switches, as core devices, are responsible for efficiently forwarding data flows. With the diversification of network applications and the increasing demand for high bandwidth, network traffic presents complex distribution characteristics, including high-frequency traffic and low-frequency traffic. In practical applications, high-frequency traffic usually occupies a large amount of bandwidth resources. In order to improve the transmission efficiency of the overall network, dynamic traffic compression technology is usually used in existing technologies to optimize high-frequency traffic. However, low-frequency traffic has real-time and bursty characteristics in some scenarios, and its priority is often ignored, which leads to a series of technical problems.

[0003] When large-scale high-frequency traffic is dynamically compressed in the prior art, the real-time requirements of low-frequency traffic are not fully considered, which may lead to the loss or reordering of low-frequency traffic, thereby affecting the reliability and stability of its transmission. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a network switch control method and system to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A network switch control method comprises the following steps: Monitor all network traffic entering the network switch, obtain the frequency characteristics of each traffic, and classify the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic; Apply dynamic traffic compression algorithm to compress high-frequency traffic; The topological links of low-frequency traffic are analyzed by sparse representation iterative reweighted least squares method to identify the interference intensity of links at different levels; The data packet sequence of low-frequency traffic is analyzed through a variational autoencoder to evaluate the logical coherence level of the traffic packets; Determine the priority scheduling strategy for low-frequency traffic based on the interference intensity of links at different levels and the logical coherence level of traffic packets; In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

[0006] In a preferred embodiment, all network traffic entering the network switch is monitored, the frequency characteristics of each traffic are obtained, and the network traffic is classified into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic, specifically including: Collect all traffic data entering the network switch in real time and record the basic characteristic information of each traffic; Extract frequency characteristics of collected flow data based on a preset frequency analysis model and output frequency characteristic values ​​of each flow; The frequency characteristic value is input into the classification rule module, and the traffic is divided into high-frequency traffic and low-frequency traffic categories according to the classification standard of the frequency range, and corresponding identification information is added to the high-frequency traffic and low-frequency traffic.

[0007] In a preferred embodiment, high-frequency traffic is compressed using a dynamic traffic compression algorithm, specifically including: Read the data classified as high-frequency traffic to determine the real-time bandwidth usage of the high-frequency traffic; Dynamically evaluate the real-time bandwidth occupancy of high-frequency traffic and generate compression parameters; A dynamic traffic compression algorithm is applied to compress high-frequency traffic data based on compression parameters.

[0008] In a preferred embodiment, the topological links of low-frequency traffic are analyzed by sparse representation iterative reweighted least squares method to identify the link interference strength at different levels, specifically including: Extract the traffic data marked as low-frequency traffic and associate it with the corresponding topological link information; Construct a sparse interference matrix containing link bandwidth utilization and traffic load distribution as analysis input; The sparse interference matrix is ​​solved using iterative reweighted least squares to identify significant interference patterns. The interference strength of links at different levels is calculated based on the significant interference patterns, and a link interference strength matrix is ​​generated.

[0009] In a preferred embodiment, the sparse interference matrix is ​​solved using iterative reweighted least squares to identify significant interference patterns, specifically: The iterative reweighted least squares method is applied to the sparse interference matrix, and the interference weight of the link is solved by the optimization function. The optimization objective function is: ;in, , represents the interference weight set of the link; For the The interference weight of the link; is the number of links; is the amount of low-frequency traffic; Indicates The low-frequency flow is Bandwidth ratio on the links; The solution process of iterative reweighted least squares method includes: Initialize the interference weight set ; Calculate the weighted error term and update Values: ;in, Indicates The link in The weight value obtained after the iteration is updated; is a regularization parameter to prevent the weight from going to zero; Indicates the number of iterations; Repeat the iteration until the weight change meets the convergence condition.

[0010] In a preferred embodiment, the data packet sequence of the low-frequency traffic is analyzed by a variational autoencoder to evaluate the logical coherence level of the traffic packet, specifically including: Extract the data packet sequence of low-frequency traffic, rearrange it in time order and construct a sequence sample set; Perform feature encoding on the sequence sample set and convert the attributes of each data packet into a high-dimensional vector representation; Use the variational autoencoder model to reduce the dimension of the feature vector and generate the potential feature representation of the sequence; Based on the latent feature representation, a logical coherence score is calculated to quantify the logical coherence level of the traffic packet.

[0011] In a preferred embodiment, based on the potential feature representation, a logical coherence score is calculated to quantify the logical coherence level of the traffic packet, specifically: Latent feature representation generated by variational autoencoder , quantify the logical coherence of each time window: ;in, represents the logical coherence score, It is The latent feature representation of the time window is represents the feature difference of adjacent windows, Indicates the total number of time windows, Indicates the number of the time window.

[0012] In a preferred embodiment, based on the interference intensity of links at different levels and the logical coherence level of traffic packets, a priority scheduling strategy for low-frequency traffic is determined, specifically including: Read the link interference intensity matrix and the logical coherence score, and calculate the resource allocation priority of low-frequency traffic on the link based on the link interference intensity matrix; Based on the logical coherence score, the priority weight of low-frequency traffic is adjusted to generate a comprehensive priority score; According to the comprehensive priority score sorting, a priority scheduling strategy for low-frequency traffic is generated.

[0013] In a preferred embodiment, in a network switch, resource allocation and scheduling are performed on compressed high-frequency traffic and prioritized low-frequency traffic, specifically including: According to the compressed high-frequency traffic and priority scheduling strategy, the resource requirements of each traffic are extracted from the scheduling table; Combined with the current available resource status of the network switch, calculate the resource allocation ratio of high-frequency traffic to low-frequency traffic; Prioritize resources required for low-frequency traffic according to the allocation ratio, and allocate the remaining resources to high-frequency traffic; Update resource allocation tables and perform scheduling operations in real time in network switches.

[0014] On the other hand, the present invention provides a network switch control system, including a traffic monitoring classification module, a high-frequency traffic compression module, a link interference analysis module, a logical coherence evaluation module, a priority strategy determination module, and a resource allocation scheduling module; Traffic monitoring and classification module: monitors all network traffic entering the network switch, obtains the frequency characteristics of each traffic, and classifies the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic; High-frequency traffic compression module: applies dynamic traffic compression algorithm to high-frequency traffic for compression processing; Link interference analysis module: Analyzes the topological links of low-frequency traffic through sparse representation iterative reweighted least squares method to identify the link interference intensity at different levels; Logical coherence evaluation module: This module analyzes the data packet sequence of low-frequency traffic through a variational autoencoder to evaluate the logical coherence level of the traffic packets. Priority strategy determination module: Determines the priority scheduling strategy for low-frequency traffic based on the interference intensity of links at different levels and the logical coherence level of traffic packets; Resource allocation and scheduling module: In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

[0015] The technical effects and advantages of a network switch control method and system of the present invention are as follows: 1. The present invention classifies network traffic and divides the traffic entering the network switch into high-frequency traffic and low-frequency traffic according to frequency characteristics, thereby achieving targeted processing. A dynamic traffic compression algorithm is applied to high-frequency traffic, which can not only effectively reduce the occupation of network bandwidth by high-frequency traffic, but also avoid the conflict problem of low-frequency traffic in network resource allocation by optimizing the compression strategy. The topological link of low-frequency traffic is analyzed by sparse representation iterative reweighted least squares method, which can accurately identify the interference intensity of different levels in the link, and provide an accurate link status basis for subsequent priority scheduling of low-frequency traffic, thereby improving the transmission stability and reliability of low-frequency traffic.

[0016] 2. By using a variational autoencoder to perform logical continuity analysis on the data packet sequence of low-frequency traffic, the present invention further quantifies the transmission characteristics of low-frequency traffic, so that the priority scheduling strategy can be dynamically optimized in combination with link interference and data packet logical characteristics. In the network switch, through real-time resource allocation and scheduling mechanism, the resource requirements of low-frequency traffic are prioritized, and at the same time, compressed high-frequency traffic resources are reasonably allocated to ensure the coordinated transmission of high-frequency and low-frequency traffic, solving the problem of high-frequency traffic compression interfering with low-frequency traffic, and improving the utilization efficiency of network resources and the real-time transmission performance of low-frequency traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a network switch control method of the present invention; Figure 2 The present invention is a schematic diagram of the structure of a network switch control system. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: Figure 1 The present invention provides a network switch control method, which comprises the following steps: Monitor all network traffic entering the network switch, obtain the frequency characteristics of each flow, and classify the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic.

[0020] A dynamic traffic compression algorithm is applied to high-frequency traffic for compression processing.

[0021] The topological links of low-frequency traffic are analyzed through sparse representation iterative reweighted least squares method to identify the interference intensity of links at different levels.

[0022] The data packet sequence of low-frequency traffic is analyzed by variational autoencoder to evaluate the logical coherence level of the traffic packets.

[0023] Based on the interference intensity of links at different levels and the logical continuity level of traffic packets, the priority scheduling strategy for low-frequency traffic is determined.

[0024] In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

[0025] Monitor all network traffic entering the network switch, obtain the frequency characteristics of each flow, and classify the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic, including: Collect all traffic data entering the network switch in real time and record the basic characteristic information of each traffic: By capturing all data packets on the switch port in real time, basic feature information such as the source address, destination address, packet size, protocol type, etc. of the traffic is recorded in sequence; each data packet is sorted by timestamp to ensure that the collected traffic data is continuous in the time dimension; the collected data is stored in the buffer area as the basis for subsequent frequency feature extraction.

[0026] Based on the preset frequency analysis model, the frequency characteristics of the collected flow data are extracted, and the frequency characteristic value of each flow is output: By reading the data in the buffer area, the data is grouped according to the source address and destination address of the traffic, and their respective time series are generated; the frequency analysis method of short-time Fourier transform is used to process each time series, and the average value of the main frequency components is calculated as the frequency characteristic value of the corresponding traffic. All extracted frequency characteristic values ​​are stored in the feature database, and unique identification information is attached to correspond to the original traffic data.

[0027] The frequency characteristic value is input into the classification rule module, and the traffic is divided into high-frequency traffic and low-frequency traffic categories according to the classification standard of the frequency range, and corresponding identification information is added to the high-frequency traffic and low-frequency traffic: Read the frequency feature value in the feature database and perform classification operations according to the set classification rules (such as the frequency of high-frequency traffic is greater than a certain threshold, and the frequency of low-frequency traffic is lower than the threshold); the classification result is associated with the unique identifier of the original traffic, and the classification result and classification basis information are marked. At the same time, a category label is added to the classified traffic for subsequent processing.

[0028] Dynamic traffic compression algorithms are used to compress high-frequency traffic, including: Read the data classified as high-frequency traffic to determine the real-time bandwidth usage of high-frequency traffic: The traffic data marked as high-frequency traffic is extracted from the classification results, and the bandwidth monitoring tool is called to perform real-time analysis on the extracted traffic to calculate the average bandwidth occupancy and peak bandwidth occupancy of each traffic. The instantaneous transmission rate and bandwidth usage pattern of each traffic are obtained by combining the size of the data packet with its transmission time. At the same time, the historical transmission records of high-frequency traffic are counted to determine the long-term trend of traffic bandwidth occupancy. All calculation results are stored in a temporary storage area to provide input for the next step of dynamic evaluation.

[0029] Dynamically evaluate the real-time bandwidth usage of high-frequency traffic and generate compression parameters: According to the stored bandwidth occupancy, a real-time optimization algorithm is used to dynamically evaluate the bandwidth usage pattern. By calculating the network load coefficient and traffic burst index, the compression ratio and redundant data discarding threshold suitable for the current network status are determined. The compression ratio is determined by the proportion of bandwidth occupancy and the remaining amount of available network resources. The redundant data discarding threshold is dynamically adjusted by detecting the data packet repetition rate. All compression parameters are stored with the unique traffic identifier as the index so that they can be directly called in subsequent compression processing.

[0030] Apply dynamic traffic compression algorithm to compress high-frequency traffic data based on compression parameters: Read the dynamically generated compression parameters and call the dynamic traffic compression algorithm to process high-frequency traffic. The compression algorithm uses redundant data deduplication technology based on content identification. First, it analyzes the content characteristics of the traffic data packet to detect and remove duplicate content. Secondly, it applies the predictive coding method to encode the content that may be reused into a short label form to further reduce the amount of data. The processed data packet maintains the temporal order of the original traffic and adds compression label information to ensure the compatibility of subsequent operations.

[0031] The topological links of low-frequency traffic are analyzed by sparse representation iterative reweighted least squares method to identify the link interference strength at different levels, including: Extract the traffic data marked as low-frequency traffic and associate it with the corresponding topological link information: Data marked as low-frequency traffic is extracted from the classification results. The extracted low-frequency traffic data includes traffic identification, source address, destination address, and time series information. At the same time, based on the topological structure information, the transmission path corresponding to the low-frequency traffic is associated with the topological link. The associated information includes the link number through which the low-frequency traffic passes, the physical connection relationship between the links, and the transmission direction. These associated information are stored as a topological link data structure, and the real-time status of each link is used as the basic input of the subsequent sparse interference matrix.

[0032] Construct a sparse interference matrix containing link bandwidth utilization and traffic load distribution as analysis input: Based on the extracted low-frequency traffic and topological link information, a sparse interference matrix is ​​constructed. Assume that the network topology contains Links and low-frequency traffic, construct a sparse interference matrix of size M×N, and each element in the sparse interference matrix is ​​expressed as: ;in, Indicates The low-frequency flow is The bandwidth ratio on the links is calculated by the traffic demand and the available bandwidth of the links; Indicates The bandwidth requirement of low-frequency traffic in bits per second; Indicates The low-frequency traffic flows through The probability of a link; Indicates The total available bandwidth of the links, in bits per second.

[0033] The sparse interference matrix only retains the elements with higher bandwidth utilization and sets the remaining elements to zero to ensure the sparsity of the matrix. The constructed matrix is ​​stored in the form of a sparse matrix to provide input for subsequent iterative solutions.

[0034] The sparse interference matrix is ​​solved using iterative reweighted least squares to identify significant interference patterns: The iterative reweighted least squares (IRLS) method is applied to the sparse interference matrix to identify the significant interference patterns in the topological links. The goal of IRLS is to solve the interference weight of the link by optimizing the function. The optimization objective function is: ;in, , represents the interference weight set of the link; For the The interference weight of the link, the initial value is 1; is the number of links; is the amount of low-frequency traffic.

[0035] IRLS solution process: Initialize the interference weight set .

[0036] Calculate the weighted error term and update Values: ;in, Indicates The link in The weight value obtained after the iteration is updated; is a regularization parameter to prevent the weight from going to zero; Indicates the number of iterations, used to track each round of the IRLS algorithm.

[0037] Repeat the iteration until the weight change meets the convergence condition.

[0038] The final Indicates the interference weight of the link. The larger the interference weight, the more significant the interference pattern of the link.

[0039] Among them, generally speaking, is an empirical value, and its size is determined by the numerical range and sparsity of the actual problem. If the numerical range of the sparse interference matrix is ​​large, a smaller value can be selected. ; If the sparse interference matrix is ​​sparse, a larger to stabilize the calculation.

[0040] Among them, the link with less interference occupies a larger value in the final weight, which means that the link has less impact on the scheduling of low-frequency traffic; the link with significant interference occupies a smaller value in the final weight, indicating that the link needs to be given priority to avoid its adverse impact on traffic.

[0041] The interference strength of links at different levels is calculated based on the significant interference pattern, and the link interference strength matrix is ​​generated: Using the obtained interference weight set, the links at different levels are grouped to calculate the interference strength. The link interference strength is defined as: ;in, Indicates The link interference strength at each level, Indicates The link set contained in each level, For the The interference weight of the link, Indicates the layer number to which the link belongs in the network topology.

[0042] The calculation result of the link interference strength is stored as a link interference strength matrix, in which each element corresponds to the interference strength of a layer.

[0043] By analyzing the topological links of low-frequency traffic through sparse representation iterative reweighted least squares method, the interference intensity of links at different levels can be efficiently identified. First, by constructing a sparse interference matrix, the bandwidth usage characteristics of low-frequency traffic in complex network topologies can be accurately reflected, while avoiding processing non-critical link data, thereby reducing computational complexity; secondly, the iterative reweighted least squares method can dynamically adjust weights and adaptively optimize significant interference patterns to improve the accuracy of the analysis results; finally, by calculating the hierarchical interference intensity of the link, a clear quantitative basis is provided for subsequent traffic priority scheduling, making resource allocation more targeted and scientific. Overall, this step takes into account both the efficiency and accuracy of the analysis and is suitable for dynamic traffic management in large-scale network environments.

[0044] The variational autoencoder is used to analyze the data packet sequence of low-frequency traffic and evaluate the logical coherence level of the traffic packets, including: Extract the packet sequence of low-frequency traffic, rearrange it in chronological order and construct a sequence sample set: The data packet sequence marked as low-frequency traffic is extracted from the classification results. The extracted attributes include source address, destination address, timestamp, protocol type and packet size. The extracted data packets are rearranged according to timestamps to generate a time-continuous sequence data set.

[0045] To ensure the integrity of the samples, the time series is divided into windows of fixed length, each containing a certain number of data packets, and the windows of insufficient length are filled with null values. The generated sequence sample set is stored in a structured form to provide input for subsequent steps.

[0046] Perform feature encoding on the sequence sample set and convert the attributes of each data packet into a high-dimensional vector representation: Perform feature encoding on the sequence sample set generated in step 1. Assume that the attribute set of the data packet includes source address, destination address, timestamp, protocol type and data packet size. Encode these attributes independently: Source address and destination address: encoded as a fixed-dimensional vector through hash mapping; Timestamp: converted into a standard time interval and mapped to the [0,1][0,1] interval using a normalization method; Protocol type: One-hot encoding is used to map discrete values ​​into vectors; Packet size: Logarithmically normalized to enhance the smoothness of the value range.

[0047] Assume that the feature vector of the data packet is represented as , then the sequence samples of the entire time window can be expressed as a feature matrix ;in, Indicates the total number of packets contained in the time window. Represents the index of a single packet in the time window in the sequence sample, It is a d-dimensional vector, where d is the feature dimension after encoding.

[0048] The variational autoencoder model is used to reduce the dimension of the feature vector and generate the potential feature representation of the sequence: For the feature matrix Input the variational autoencoder model for dimensionality reduction. The core of the variational autoencoder consists of an encoder and a decoder, and its goal is to learn the potential feature representation .

[0049] The encoder transforms Mapped to distribution parameters in latent space: , ;in, is the mean of the underlying distribution, is the standard deviation of the underlying distribution, and It is a trainable neural network.

[0050] in, Represents the encoder neural network used to calculate the mean of the potential distribution in the variational autoencoder, which maps the input high-dimensional feature matrix through multiple layers of nonlinear transformations to ; represents the encoder neural network used to calculate the standard deviation of the latent distribution in the variational autoencoder, which is similar to , but its goal is to output ; Represents the central position in the potential feature distribution and is the encoder network The output of is used to describe the principal component positions of the potential distribution in the latent space; represents the extended range of the potential feature distribution, which is the encoder network The output of is used to describe the degree of variation of features in the latent space. The larger the standard deviation, the more dispersed the distribution.

[0051] Sampling from a Gaussian distribution using the reparameterization trick: ;in, is the potential feature representation after dimensionality reduction, is the noise sampled from a standard normal distribution, represents the standard normal distribution.

[0052] The decoder starts from the latent representation Reconstructing input features : ;in, is the reconstructed feature matrix, is the decoder network.

[0053] in, Represents the decoder neural network in the variational autoencoder. It uses multiple layers of nonlinear mapping to represent the potential features. Restore the input features to high-dimensional space ,The goal of the decoder is to reconstruct the input features while minimizing the error with the original input.

[0054] The training goal is to minimize the reconstruction error and the KL divergence of the latent distribution: ;in, represents the total loss function of the variational autoencoder; Represents the potential distribution With prior distribution The KL divergence of is used to constrain the distribution shape of the latent space to make it close to the standard normal distribution.

[0055] Based on the latent feature representation, a logical coherence score is calculated to quantify the logical coherence level of the traffic packet: Latent feature representation generated by variational autoencoder , quantify the logical coherence of each time window. The specific calculation formula is as follows: ;in, represents the logical coherence score, It is The latent feature representation of the time window is represents the feature difference of adjacent windows, Indicates the total number of time windows, Indicates the number of the time window.

[0056] The lower the logical coherence score, the higher the logical coherence of the traffic packet; the higher the logical coherence score, the worse the logical coherence; the final result is stored as a logical coherence score matrix for subsequent priority scheduling strategies.

[0057] Based on the interference intensity of links at different levels and the logical coherence level of traffic packets, the priority scheduling strategy for low-frequency traffic is determined, including: Read the link interference strength matrix and logical coherence score, and calculate the resource allocation priority of low-frequency traffic on the link based on the link interference strength matrix: The link interference intensity related to low-frequency traffic is extracted from the link interference intensity matrix. The link interference intensity of each link indicates the congestion or stability of the link. Then, for each low-frequency traffic, the link interference intensity of all links on the path it passes through is analyzed, and according to the distribution of traffic on each link, its overall resource allocation priority is calculated. The calculation method is to comprehensively consider the occupancy ratio of traffic on each link and the cumulative result of the corresponding link interference intensity. The final result is stored in the form of a priority allocation table for subsequent analysis.

[0058] Based on the logical coherence score, the priority weight of low-frequency traffic is adjusted to generate a comprehensive priority score: The logical coherence score of each low-frequency flow is read. The logical coherence score reflects the integrity level of the flow packet sequence during transmission. The logical coherence score is used as a weight factor to adjust the impact of the link resource allocation priority. The specific adjustment method is: give a higher adjustment factor weight to the flow with poor logical coherence to reduce its overall priority; for the flow with good logical coherence, retain its original impact value of the link allocation priority, and the adjusted comprehensive priority score is recorded in the priority score table as the basis for generating the scheduling strategy.

[0059] The comprehensive priority score is the adjusted result calculated by combining the link resource priority and the logical coherence score. The specific calculation logic is as follows: The lower the logical coherence score, the greater the impact on the link resource priority; the higher the logical coherence score, the lower the weight of the link priority.

[0060] The calculation of the comprehensive priority score can be described as: the link priority allocation value is weighted and modified according to the proportion of the logical coherence score, and the modified value is the comprehensive priority score. The correction rule is determined by the logical coherence score and a balance factor to ensure that the impact of the link priority and the logical coherence score reaches a reasonable trade-off in the overall scheduling.

[0061] According to the comprehensive priority score sorting, a priority scheduling strategy for low-frequency traffic is generated: The comprehensive priority scores of low-frequency traffic are sorted, and a priority scheduling table is generated in descending order of scores. The scheduling table contains the priority level of each low-frequency traffic, the corresponding link path, and the scheduling rules. High-priority traffic is allocated more network resources, and the scheduling rules ensure that it is processed before low-priority traffic. The priority scheduling table can be provided to network switches for real-time scheduling.

[0062] In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic, including: According to the compressed high-frequency traffic and priority scheduling strategy, the resource requirements of each traffic are extracted from the scheduling table: Read the compressed high-frequency traffic and low-frequency traffic priority scheduling table information. The scheduling table includes the bandwidth requirements, delay requirements, priority level and corresponding link path data of each traffic.

[0063] The resource requirements of each flow are analyzed to extract the actual bandwidth requirement value and the minimum guaranteed bandwidth threshold. This information provides a basis for resource allocation and ensures that the requirements of high-frequency and low-frequency flows can be accurately identified.

[0064] Combined with the current available resource status of the network switch, calculate the resource allocation ratio of high-frequency traffic to low-frequency traffic: By querying the resource usage of the current link, including the remaining bandwidth, link load, and available time slices, the resource allocation ratio is calculated according to the resource requirements of high-frequency traffic and low-frequency traffic. The resource allocation ratio gives priority to meeting the needs of low-frequency traffic, and dynamically adjusts the allocation strategy of the remaining resources according to the compressed bandwidth requirements of high-frequency traffic. The allocation ratio comprehensively considers the priority of the traffic, the link load, and the fairness of bandwidth allocation, and the results are recorded as an allocation ratio table.

[0065] Prioritize resources required for low-frequency traffic according to the allocation ratio, and allocate the remaining resources to high-frequency traffic: Based on the allocation ratio table, resources are first allocated to low-frequency traffic to ensure that its key requirements are met, such as the minimum guaranteed bandwidth and priority of the link. Then, the remaining link resources are allocated to high-frequency traffic, using its compressed traffic characteristics to complete the transmission with the least possible resource occupation.

[0066] During the allocation process, the allocation status of each link is monitored in real time to avoid resource allocation exceeding the available capacity of the link.

[0067] Update resource allocation tables and perform scheduling operations in real time in network switches: The allocated resource information is recorded in the resource allocation table, including the allocated bandwidth, priority level and corresponding link resource usage of each flow; the real-time updated allocation table is used to guide the execution of scheduling rules of the network switch.

[0068] During the scheduling process, link status changes are monitored. If resources are insufficient or the load exceeds the limit, reallocation operations are triggered to ensure the continuity and stability of resource allocation and scheduling.

[0069] Embodiment 2: The difference between Embodiment 2 of the present invention and Embodiment 1 is that this embodiment introduces a network switch control system.

[0070] Figure 2 A structural schematic diagram of a network switch control system of the present invention is given, which includes a traffic monitoring classification module, a high-frequency traffic compression module, a link interference analysis module, a logical coherence evaluation module, a priority strategy determination module and a resource allocation scheduling module.

[0071] Traffic monitoring and classification module: monitors all network traffic entering the network switch, obtains the frequency characteristics of each traffic, and classifies the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic.

[0072] High-frequency traffic compression module: applies dynamic traffic compression algorithm to high-frequency traffic for compression processing.

[0073] Link interference analysis module: Analyzes the topological links of low-frequency traffic through sparse representation iterative reweighted least squares method to identify the link interference intensity at different levels.

[0074] Logical coherence evaluation module: The data packet sequence of low-frequency traffic is analyzed through a variational autoencoder to evaluate the logical coherence level of the traffic packet.

[0075] Priority strategy determination module: Determines the priority scheduling strategy for low-frequency traffic based on the interference intensity of links at different levels and the logical continuity level of traffic packets.

[0076] Resource allocation and scheduling module: In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

[0077] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0078] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0079] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0081] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0082] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0083] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0084] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for 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 methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0085] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0086] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A network switch control method, characterized in that: The steps include: Monitor all network traffic entering the network switch, obtain the frequency characteristics of each traffic, and classify the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic; Apply dynamic traffic compression algorithm to compress high-frequency traffic; The topological links of low-frequency traffic are analyzed by sparse representation iterative reweighted least squares method to identify the interference intensity of links at different levels; The data packet sequence of low-frequency traffic is analyzed through a variational autoencoder to evaluate the logical coherence level of the traffic packets; Determine the priority scheduling strategy for low-frequency traffic based on the interference intensity of links at different levels and the logical coherence level of traffic packets; In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

2. A network switch control method according to claim 1, characterized in that: Monitor all network traffic entering the network switch, obtain the frequency characteristics of each flow, and classify the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic, including: Collect all traffic data entering the network switch in real time and record the basic characteristic information of each traffic; Extract frequency characteristics of collected flow data based on a preset frequency analysis model and output frequency characteristic values ​​of each flow; The frequency characteristic value is input into the classification rule module, and the traffic is divided into high-frequency traffic and low-frequency traffic categories according to the classification standard of the frequency range, and corresponding identification information is added to the high-frequency traffic and low-frequency traffic.

3. A network switch control method according to claim 2, characterized in that: Dynamic traffic compression algorithms are used to compress high-frequency traffic, including: Read the data classified as high-frequency traffic to determine the real-time bandwidth usage of the high-frequency traffic; Dynamically evaluate the real-time bandwidth occupancy of high-frequency traffic and generate compression parameters; A dynamic traffic compression algorithm is applied to compress high-frequency traffic data based on compression parameters.

4. A network switch control method according to claim 3, characterized in that: The topological links of low-frequency traffic are analyzed by sparse representation iterative reweighted least squares method to identify the link interference strength at different levels, including: Extract the traffic data marked as low-frequency traffic and associate it with the corresponding topological link information; Construct a sparse interference matrix containing link bandwidth utilization and traffic load distribution as analysis input; The sparse interference matrix is ​​solved using iterative reweighted least squares to identify significant interference patterns. The interference strength of links at different levels is calculated based on the significant interference patterns, and a link interference strength matrix is ​​generated.

5. A network switch control method according to claim 4, characterized in that: The sparse interference matrix is ​​solved by iterative reweighted least squares method to identify significant interference patterns, specifically: The iterative reweighted least squares method is applied to the sparse interference matrix, and the interference weight of the link is solved by the optimization function. The optimization objective function is: ;in, , represents the interference weight set of the link; For the The interference weight of the link; is the number of links; is the amount of low-frequency traffic; Indicates The low-frequency flow is Bandwidth ratio on the links; The solution process of iterative reweighted least squares method includes: Initialize the interference weight set ; Calculate the weighted error term and update Values: ;in, Indicates The link in The weight value obtained after the iteration is updated; is a regularization parameter to prevent the weight from going to zero; Indicates the number of iterations; Repeat the iteration until the weight change meets the convergence condition.

6. A network switch control method according to claim 5, characterized in that: The variational autoencoder is used to analyze the data packet sequence of low-frequency traffic and evaluate the logical coherence level of the traffic packets, including: Extract the data packet sequence of low-frequency traffic, rearrange it in time order and construct a sequence sample set; Perform feature encoding on the sequence sample set and convert the attributes of each data packet into a high-dimensional vector representation; Use the variational autoencoder model to reduce the dimension of the feature vector and generate the potential feature representation of the sequence; Based on the latent feature representation, a logical coherence score is calculated to quantify the logical coherence level of the traffic packet.

7. A network switch control method according to claim 6, characterized in that: Based on the latent feature representation, the logical coherence score is calculated to quantify the logical coherence level of the traffic packet, specifically: Latent feature representation generated by variational autoencoder , quantify the logical coherence of each time window: ;in, represents the logical coherence score, It is The latent feature representation of the time window is represents the feature difference of adjacent windows, Indicates the total number of time windows, Indicates the number of the time window.

8. A network switch control method according to claim 7, characterized in that: Based on the interference intensity of links at different levels and the logical coherence level of traffic packets, the priority scheduling strategy for low-frequency traffic is determined, including: Read the link interference intensity matrix and the logical coherence score, and calculate the resource allocation priority of low-frequency traffic on the link based on the link interference intensity matrix; Based on the logical coherence score, the priority weight of low-frequency traffic is adjusted to generate a comprehensive priority score; According to the comprehensive priority score sorting, a priority scheduling strategy for low-frequency traffic is generated.

9. A network switch control method according to claim 8, characterized in that: In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic, including: According to the compressed high-frequency traffic and priority scheduling strategy, the resource requirements of each traffic are extracted from the scheduling table; Combined with the current available resource status of the network switch, calculate the resource allocation ratio of high-frequency traffic to low-frequency traffic; Prioritize resources required for low-frequency traffic according to the allocation ratio, and allocate the remaining resources to high-frequency traffic; Update resource allocation tables and perform scheduling operations in real time in network switches.

10. A network switch control system, used to implement a network switch control method according to any one of claims 1 to 9, characterized in that: It includes traffic monitoring and classification module, high-frequency traffic compression module, link interference analysis module, logical coherence evaluation module, priority strategy determination module and resource allocation and scheduling module; Traffic monitoring and classification module: monitors all network traffic entering the network switch, obtains the frequency characteristics of each traffic, and classifies the network traffic into high-frequency traffic and low-frequency traffic according to the frequency characteristics of the traffic; High-frequency traffic compression module: applies dynamic traffic compression algorithm to high-frequency traffic for compression processing; Link interference analysis module: Analyzes the topological links of low-frequency traffic through sparse representation iterative reweighted least squares method to identify the link interference intensity at different levels; Logical coherence evaluation module: This module analyzes the data packet sequence of low-frequency traffic through a variational autoencoder to evaluate the logical coherence level of the traffic packets. Priority strategy determination module: Determines the priority scheduling strategy for low-frequency traffic based on the interference intensity of links at different levels and the logical coherence level of traffic packets; Resource allocation and scheduling module: In the network switch, resources are allocated and scheduled for compressed high-frequency traffic and prioritized low-frequency traffic.

Images