Network path switching method and device based on multi-cloud network and computer equipment

Abstract

The invention relates to a network path switching method and device based on a multi-cloud network and computer equipment. The method comprises the following steps: performing trend prediction processing based on a network performance data set of each network path obtained in advance to obtain change trend data; performing feature extraction processing and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction; and executing a network path switching action based on the network path switching instruction, so that a transmission path of service data is switched from a current network path to a target network path. By adopting the method, the working efficiency of the cloud network can be improved.

Description

Technical Field

[0001] The present application relates to the field of cloud computing technology, and in particular to a network path switching method, device and computer equipment based on a multi-cloud network. Background Art

[0002] In the digital age, the rapid development of information technology has profoundly changed the operating model of enterprises and organizations. With its powerful computing power, flexible resource allocation, and efficient data storage and processing advantages, cloud computing has become the core technical support for promoting enterprises to enhance their competitiveness and promote innovative development. With the continuous expansion of business and the sharp increase in data volume, enterprises and organizations are increasingly dependent on cloud computing.

[0003] In order to reduce risks, improve service quality, and give full play to the features of different cloud service providers, more and more enterprises are choosing to adopt a multi-cloud environment as their technology deployment model. This model enables enterprises to flexibly select and combine multiple cloud services according to specific needs, thereby achieving optimal resource allocation and ensuring the continuous and stable operation of the business.

[0004] However, although the multi-cloud environment has brought more flexibility and improved resource utilization efficiency to enterprises, in practical applications, related network connection methods still face many challenges. Existing methods usually lack efficient collaboration and information interaction between modules in the multi-cloud network, which leads to poor data flow and delayed information sharing, thus affecting the working efficiency of the multi-cloud network. Summary of the invention

[0005] Based on this, it is necessary to provide a network path switching method, device and computer equipment based on a multi-cloud network that can improve the working efficiency of the multi-cloud network in response to the above technical problems.

[0006] In a first aspect, the present application provides a network path switching method based on a multi-cloud network, comprising:

[0007] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0008] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0009] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0010] In one embodiment, the network performance data set includes a plurality of sub-data sets arranged in chronological order, and trend prediction processing is performed based on the pre-acquired network performance data sets of each network path to obtain change trend data, including:

[0011] Iterative calculation is performed for each sub-data set. In the first iteration process, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend features of the first iteration;

[0012] In the non-first iteration process, the last change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the sub-data set of this iteration and the last change trend feature are sequentially subjected to feature extraction and feature fusion processing by the trend prediction model to obtain the change trend feature of this iteration;

[0013] The change trend feature obtained based on the last sub-data set is determined as the change trend data.

[0014] In one embodiment, the network performance data set and the change trend data are sequentially subjected to feature extraction and adjustment processing to obtain a network path switching instruction, including:

[0015] The instruction generation model is used to perform feature extraction and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features;

[0016] Adjust the fused data features based on the network optimization target to obtain adjusted fused data features;

[0017] A network path switching instruction is generated based on the fused data features and the adjusted fused data features.

[0018] In one embodiment, the instruction generation model is used to perform feature extraction and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features, including:

[0019] Using the instruction generation model to perform feature extraction processing on the network performance data set, the performance characteristics are obtained;

[0020] The instruction generation model is used to perform feature extraction processing on the change trend data to obtain data features;

[0021] The performance features and data features are fused to obtain fused data features.

[0022] In one embodiment, the method further comprises:

[0023] Based on the fused data features, determine the joint loss function of the trend prediction model and the instruction generation model;

[0024] The trend prediction model and the instruction generation model are adjusted based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0025] In one embodiment, the method further comprises:

[0026] The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

[0027] In a second aspect, the present application also provides a network path switching device based on a multi-cloud network, comprising:

[0028] A prediction processing module is used to perform trend prediction processing based on a set of network performance data of each network path acquired in advance to obtain change trend data;

[0029] An adjustment processing module is used to perform feature extraction and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction;

[0030] The path switching module is used to execute a network path switching action based on a network path switching instruction so that the transmission path of the service data is switched from the current network path to the target network path.

[0031] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0032] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0033] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0034] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0036] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0037] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0038] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0039] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0040] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0041] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0042] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0043] The above-mentioned network path switching method, device and computer equipment based on multi-cloud network first perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data; then perform feature extraction processing and adjustment processing on the network performance data set and the change trend data in turn to obtain a network path switching instruction; finally, execute the network path switching action based on the network path switching instruction to switch the transmission path of the business data from the current network path to the target network path; in this method, by generating a network path switching instruction based on the network performance data set and the trend prediction data, dynamic path optimization can be performed according to the real-time status and future trends of the network, thereby helping to reduce network congestion, optimize resource allocation, and thus improve the working efficiency of the multi-cloud network. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 is an internal structure diagram of a computer device in one embodiment;

[0046] Figure 2 It is a flowchart of a network path switching method based on a multi-cloud network in one embodiment;

[0047] Figure 3 is a flow chart of a network path switching method based on a multi-cloud network in another embodiment;

[0048] Figure 4 is a flow chart of a network path switching method based on a multi-cloud network in another embodiment;

[0049] Figure 5 is a flow chart of a network path switching method based on a multi-cloud network in another embodiment;

[0050] Figure 6 is a flow chart of a network path switching method based on a multi-cloud network in another embodiment;

[0051] Figure 7 It is a structural block diagram of a network path switching device based on a multi-cloud network in another embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 1 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data in the network path switching process based on a multi-cloud network. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a network path switching method based on a multi-cloud network is implemented.

[0054] Those skilled in the art will understand that Figure 1The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0055] In an exemplary embodiment, Figure 2 As shown, a network path switching method based on a multi-cloud network is provided, and the method is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps 201 to 203. Among them:

[0056] Step 201 , performing trend prediction processing based on a set of network performance data of each network path acquired in advance to obtain change trend data.

[0057] The network path refers to the transmission path of data from the source node to the target node. Business data is transmitted through different physical or virtual links to form multiple available network paths. Due to the influence of factors such as bandwidth, delay, and packet loss rate, the network performance of each network path is different.

[0058] Network performance data sets refer to the measurement results of the actual performance indicators of network paths, including but not limited to bandwidth (data transmission rate), latency (data transmission time delay), packet loss rate (data loss ratio), etc. By monitoring network performance data sets, the health of each network path can be evaluated.

[0059] Change trend data refers to the future change trend of network path performance predicted based on the network performance data set. Based on the change trend data, it is possible to predict the performance changes that may occur in the network path in the future, such as bandwidth reduction, delay increase, etc.

[0060] In an embodiment of the present application, a computer device may perform trend prediction processing based on a set of network performance data of each network path acquired in advance to obtain change trend data.

[0061] In some embodiments, first, the computer device collects network performance data from each network path in real time. Since the collected data may contain noise or missing values, data preprocessing is required after collection, such as removing outliers, filling missing data, smoothing, etc., to ensure the accuracy and completeness of the data.

[0062] Next, data analysis and feature extraction are required. In this process, valuable features need to be extracted from the processed network performance data. These features can include the mean, maximum, minimum, and fluctuation range of bandwidth, which can reflect the overall health of the network path. In addition, computer equipment can also analyze the time series characteristics of the data, such as trends and periodic changes, to identify potential change patterns in network performance.

[0063] After feature extraction, a suitable prediction model can be selected based on the extracted features to generate the change trend of each network path in the future, that is, the change trend data.

[0064] In some embodiments, the computer device will visualize the change trend data to facilitate understanding and decision-making. The prediction results can be intuitively displayed in the form of charts, curves, or dashboards to help users quickly identify potential network bottlenecks or problems. At the same time, a detailed analysis report will be generated to summarize the future change trend of the network path and provide a scientific basis for the next step of network path switching.

[0065] Step 202, perform feature extraction processing and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction.

[0066] Feature extraction refers to the process of extracting key features from network performance data sets and change trend data. The extracted features may include network path stability, peak latency, transmission volatility, etc.

[0067] Adjustment processing refers to the process of processing and optimizing the extracted feature data. It can include steps such as data standardization and normalization to eliminate the impact of different data dimensions, or adjust the data deviation by weighting and screening the most important features to improve the accuracy of decision-making.

[0068] The network path switching instruction is a control instruction generated based on the feature extraction and adjustment processing results. It specifies when and how to switch the current network path to another target network path.

[0069] In an embodiment of the present application, a computer device performs feature extraction processing and adjustment processing on a network performance data set and a change trend data in sequence to obtain a network path switching instruction.

[0070] In some embodiments, the computer device needs to perform feature extraction on the network performance data set. At this stage, the computer device extracts key features from the performance data of each network path, such as bandwidth, delay, packet loss rate, delay jitter, etc. These features help to evaluate the current status of each network path. For example, by analyzing the average value of bandwidth and fluctuations in delay, the computer device can determine whether the current network path is stable and whether it is suitable to continue to serve as a data transmission channel. In addition, the computer device also needs to extract long-term trends and short-term fluctuation patterns from the change trend data. This information can help predict future performance changes in the network path, thereby providing support for path switching decisions.

[0071] After the features are extracted, the next step is feature adjustment processing. The goal of this stage is to optimize the feature data so that they are more suitable for subsequent decision making. Common processing methods include standardization or normalization, which can eliminate the dimensional differences between different features so that they have the same measurement scale. In addition, some features may be more important in path switching decisions, so the influence of these features can be highlighted by weighting. For example, delay and packet loss rate are more important for real-time communication, so they may need to be given higher weights. Feature selection techniques can also be used to remove redundant or irrelevant features, thereby simplifying the data and improving decision-making efficiency.

[0072] After completing feature extraction and adjustment processing, the computer device can build a decision model based on the processed feature data to determine whether a network path switch is required. The decision model can make basic path selection decisions through threshold judgments, such as triggering a switch when the delay or packet loss rate exceeds a preset tolerance value. In more complex cases, the computer device can use machine learning models (such as decision trees, support vector machines, etc.) to analyze historical data and learn the performance patterns of network paths, thereby making more accurate predictions and decisions.

[0073] Finally, based on the output of the decision model, the computer device generates a network path switching instruction. The path switching instruction specifies when and where to perform path switching to ensure the smooth transmission of business data. When executing path switching, the computer device adjusts the routing, load balancing and other configurations according to the instructions, so as to switch the data flow from the current network path to the target network path.

[0074] Step 203: Execute a network path switching action based on the network path switching instruction, so that the transmission path of the service data is switched from the current network path to the target network path.

[0075] Among them, the network path switching action refers to the process of actually executing the network path switching according to the network path switching instruction, which may include adjusting the route, reconfiguring the connection, switching the load and other operations on the network equipment, with the purpose of transferring the business data from the current network path to a better target path.

[0076] Business data refers to the data flow that needs to be transmitted in the network, usually representing the transmission content of a specific application, such as web page access data, file downloads, video streaming, etc.

[0077] In an embodiment of the present application, the computer device executes a network path switching action based on a network path switching instruction so that the transmission path of the business data is switched from the current network path to the target network path.

[0078] In some embodiments, after receiving the path switching instruction, the computer device needs to parse and verify the instruction to determine the validity and accuracy of the network path switching instruction, wherein the verification process may include verifying whether the target network path is available and whether the switching conditions have been met. For example, the target path, switching timing, priority and other information contained in the network path switching instruction will be carefully checked to avoid network interruption or instability caused by instruction errors.

[0079] Next, the computer equipment will evaluate the availability and stability of the target network path. In this process, it is necessary to check the current network performance indicators of the target network path, such as bandwidth, latency, packet loss rate, etc., to ensure that the target path can meet the transmission requirements of business data.

[0080] When the target network path is confirmed to be available, the computer device will perform the necessary resource scheduling and routing configuration. This includes updating the routing table, adjusting the routing rules on the network device, and possibly reconfiguring the load balancing policy to ensure that the data flow can smoothly switch from the current path to the target path. In addition, other related devices in the network (such as firewalls, routers, or load balancers) may also need to adjust their settings to adapt to the new path configuration.

[0081] The path switching action is formally executed after all configurations are completed. At this point, the computer device will forward the service data to the target network path according to the new routing rules. After the path switching is completed, the computer device will monitor the performance of the new target path in real time to verify the switching effect. This can include checking the bandwidth, delay, packet loss rate and other key performance indicators of the target path to ensure that the switched target network path meets the expected performance standards. If the monitoring results show that the performance of the target network path is lower than expected, the computer device should take emergency measures in a timely manner, such as falling back to the original path or reselecting other available paths.

[0082] In the above-mentioned network path switching method based on multi-cloud network, trend prediction processing is first performed based on the network performance data set of each network path acquired in advance to obtain change trend data; then, feature extraction processing and adjustment processing are performed on the network performance data set and the change trend data in turn to obtain a network path switching instruction; finally, a network path switching action is executed based on the network path switching instruction to switch the transmission path of the business data from the current network path to the target network path; in this method, by generating a network path switching instruction based on the network performance data set and trend prediction data, dynamic path optimization can be performed according to the real-time status and future trends of the network, thereby helping to reduce network congestion, optimize resource allocation, and thus improve the working efficiency of the multi-cloud network.

[0083] In an exemplary embodiment, Figure 3 As shown, the above network performance data set includes multiple sub-data sets arranged in chronological order. On this basis, the above "performing trend prediction processing based on the network performance data set of each network path obtained in advance to obtain change trend data" includes steps 301 to 303. Among them:

[0084] Step 301, iterative calculation is performed on each sub-data set. In the first iteration, the first sub-data set is input into the trend prediction model, and feature extraction is performed on the first sub-data set using the trend prediction model to obtain the change trend features of the first iteration.

[0085] Among them, a sub-data set refers to a subset divided from the overall data. Usually, these sub-data sets are divided by time and represent network performance data in a specific period. Each sub-data set can contain network performance indicators such as bandwidth, delay, and packet loss rate, which are used to analyze network change trends.

[0086] The trend prediction model is a model used to predict the future trend of data changes. By analyzing each sub-data set, the model can discover patterns and laws in the data and predict future changes based on these laws.

[0087] Feature extraction refers to extracting information useful for problem analysis from each sub-data set. These features are usually representative or key information in the data, such as bandwidth fluctuations, periodic changes in latency, etc. In trend prediction, the purpose of feature extraction is to convert complex sub-data sets into concise representations that can be used for modeling and prediction.

[0088] The change trend feature is a feature extracted from each sub-data set that describes the data change pattern. In network performance analysis, the change trend feature may include the rising or falling trend of bandwidth, the increase or decrease trend of latency, etc. These features are used to reflect the change trend of each sub-data set over time.

[0089] In the embodiment of the present application, during the first iteration, the computer device inputs the first sub-data set into the trend prediction model for processing. The first sub-data set contains preliminary performance data or other relevant data of the network path. The trend prediction model performs feature extraction processing on the first sub-data set, extracts key change trend features from it, and obtains the change trend features of the first iteration.

[0090] In some embodiments, the trend prediction model can use a recurrent neural network (RNN) because RNN can remember previous information and combine it with the current input to capture the time dependency in the data. Therefore, it has unique advantages when there are multiple sub-data sets arranged in chronological order. Assume that the input time series is , the hidden state of the RNN The update formula is as follows:

[0091]

[0092] Among them, W is the weight matrix, b is the bias term, is the activation function, is the hidden state at the previous moment, Represents the input data of the recurrent neural network at time t. During the training process, RNN adjusts weights and biases through the back-propagation algorithm to minimize the error between the predicted value and the actual value. The loss function usually adopts mean square error (MSE) or cross entropy. For long sequence data, there may be problems of gradient vanishing or gradient explosion. The RNN structure improved by using long short-term memory network (LSTM) can alleviate gradient explosion. Taking LSTM as an example, it introduces input gate, forget gate and output gate to control the flow and memory of information. Input gate Determines how much of the currently entered information is saved:

[0093]

[0094] in, is the output of the input gate, Activation function, usually the sigmoid function, is the weight matrix, is the bias term of the input gate.

[0095] Forget Gate Determines how much of the past is forgotten:

[0096]

[0097] in, is the output of the forget gate, is the weight matrix, is the bias term of the forget gate.

[0098] Output Gate Determines how much of the past is forgotten:

[0099]

[0100] in, is the output of the output gate, is the weight matrix, is the bias term for the output gate.

[0101] Through these gating mechanisms, LSTM can better handle long-term dependencies in long sequence data, thereby more accurately predicting the changing trend data of network traffic.

[0102] Step 302, in a non-first iteration process, the previous change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing and feature fusion processing on the sub-data set of this iteration and the previous change trend feature in sequence to obtain the change trend feature of this iteration.

[0103] The non-first iteration process refers to each round of calculation process after the first iteration. Unlike the first iteration, the non-first iteration not only relies on the current sub-data set during calculation, but also combines the change trend characteristics obtained from the previous round of iteration calculation to make the data more accurately analyzed and predicted.

[0104] Feature fusion processing refers to the process of merging or integrating data features from different sources or different time points. In non-first iterations, the computer device will combine the change trend features of the previous iteration with the features of the sub-data set of the current iteration. The purpose of feature fusion is to enhance the understanding of data changes by combining different features to obtain more comprehensive and accurate trend features.

[0105] In the embodiment of the present application, when entering the non-first iteration process, the computer device will perform further calculations according to the results of the previous iteration at each iteration. In this process, the computer device inputs the change trend characteristics of the previous round (i.e., the output characteristics of the previous iteration) and the new sub-data set of this iteration into the trend prediction model.

[0106] The trend prediction model performs feature extraction processing on the sub-data set of this iteration to extract the main change trend features of the current data. Then, the computer device performs feature fusion processing on the change trend features of the previous round and the features extracted this time to obtain the change trend features of this iteration.

[0107] Step 303: Determine the change trend feature obtained based on the last sub-data set as the change trend data.

[0108] In the embodiment of the present application, after several rounds of iterations, the computer device will eventually determine the overall change trend data based on the change trend characteristics obtained from the last sub-data set.

[0109] In some embodiments, the process of acquiring a set of network performance data for each network path may include: using the Simple Network Management Protocol (SNMP) to obtain a set of network performance data from the network, including port traffic, error rate, etc. The performance indicator value corresponding to a specific object identifier) ​​can be obtained through the Get and GetNext operation commands of SNMP. In addition, in combination with the sampled flow (sFlow) technology, an in-depth analysis of the traffic in the network is performed. Through the analysis, rich information such as the source address, destination address, protocol type, application type, etc. of the traffic can be obtained, which helps to more accurately evaluate network performance. By combining the SNMP and sFlow technologies, the performance parameters of the network path can be obtained comprehensively, in real time, and accurately, providing strong support for subsequent analysis, decision-making, and optimization.

[0110] In the above embodiment, in the first iteration process, the first sub-data set is used for feature extraction to generate preliminary change trend features. In the subsequent iteration process, by fusing the change trend features obtained in the previous iteration with the current sub-data set, the trend prediction can be continuously adjusted and optimized. This iterative process effectively utilizes historical data and new data, gradually improves the accuracy of trend prediction, and enables the trend prediction model to adapt to changes in network status and provide more accurate trend prediction.

[0111] In an exemplary embodiment, Figure 4 As shown, the above-mentioned "performing feature extraction and adjustment processing on the network performance data set and the change trend data in sequence to obtain the network path switching instruction" includes steps 401 to 403. Among them:

[0112] Step 401, using the instruction generation model to perform feature extraction processing and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features.

[0113] Among them, the instruction generation model is an algorithm model based on data-driven, and its main function is to generate network path switching instructions based on the input network performance data and change trend data to execute path switching operations.

[0114] Fusion data features refer to comprehensive features obtained after feature extraction and feature fusion processing. Fusion data features integrate key information of network performance data and change trend data, reflecting the current status of the network path and future change trends. Fusion data features provide comprehensive input for subsequent network path switching decisions, helping to determine whether to switch paths or select the best path.

[0115] In an embodiment of the present application, the network performance data set and the change trend data are first merged into one data set, and then the instruction generation model data set is used to perform feature extraction processing to extract relevant features therefrom, and then multiple relevant features are subjected to feature fusion processing to obtain fused data features.

[0116] Step 402: Adjust the fused data features based on the network optimization target to obtain adjusted fused data features.

[0117] The network optimization goal refers to the optimization effect that is expected to be achieved during network path switching or network resource configuration. Network optimization goals may include improving network throughput, reducing latency, reducing packet loss, optimizing bandwidth allocation, and reducing energy consumption. The determination of network optimization goals is the basis for network path switching decisions, and different optimization goals will lead to different path switching decisions.

[0118] Adjustment processing refers to the process of adjusting the fused data features based on the network optimization goal. According to the specific optimization goal, the fused data features are further optimized so that they can better guide subsequent decisions. For example, under the optimization goal of improving throughput, the adjusted fused data features may pay more attention to the correlation between bandwidth and traffic, while under the goal of reducing latency, the adjusted features may pay more attention to the network's latency characteristics.

[0119] In the embodiment of the present application, after obtaining the fused data features, the computer device will adjust and process the fused data features based on the network optimization goal. During the adjustment process, the computer device will weight or filter the fused data features according to the preset optimization goal so that the fused data features finally obtained can most effectively support the realization of the network optimization goal.

[0120] Step 403: Generate a network path switching instruction based on the fused data features and the adjusted fused data features.

[0121] In an embodiment of the present application, based on the adjusted fusion data features, the instruction generation model will generate corresponding network path switching instructions to perform the path switching operation at an appropriate time.

[0122] In the above embodiment, by extracting and fusing features of network performance data and change trend data, adjusting the data in combination with network optimization goals, and finally generating network path switching instructions, automatic network optimization and intelligent scheduling are achieved, which helps business data dynamically adapt to network changes, so that the network can achieve optimal performance in different usage scenarios.

[0123] In an exemplary embodiment, Figure 5 As shown, the above-mentioned "using the instruction generation model to perform feature extraction processing and feature fusion processing on the network performance data set and the change trend data respectively to obtain fusion data features" includes steps 501 to 503. Among them:

[0124] Step 501, using the instruction generation model to perform feature extraction processing on the network performance data set to obtain performance characteristics.

[0125] Among them, performance characteristics are key information extracted from a collection of network performance data, usually including various indicators representing the network operation status, such as bandwidth, latency, packet loss rate, throughput, etc. These characteristics reflect the performance of the network in a specific time period and are the basis for network optimization decisions.

[0126] In the embodiment of the present application, the instruction generation model extracts performance characteristics from the network performance data set, such as network bandwidth, delay, packet loss rate, throughput, etc. The performance characteristics can reflect the current operation status of the network and provide a basis for subsequent analysis.

[0127] Step 502: Use the instruction generation model to perform feature extraction processing on the change trend data to obtain data features.

[0128] Among them, data features are key information extracted from the change trend data, usually including the change trend, change amplitude, periodicity, etc. of network performance. Data features can reflect the possible future change direction of the network, such as predicting the increase or decrease trend of latency, bandwidth fluctuation, etc., thus providing a basis for decisions such as network path switching.

[0129] In the embodiment of the present application, the instruction generation model performs feature extraction processing on the change trend data to extract data features, such as bandwidth growth trend, delay change trend, traffic fluctuation, etc. The data features can reflect the future change pattern of the network.

[0130] Step 503: perform feature fusion processing on the performance features and the data features to obtain fused data features.

[0131] Among them, the fused data features are a comprehensive feature set obtained through feature extraction and fusion processing. The fused data features contain key information about the current network performance (such as bandwidth, latency, throughput, etc.) and change trends (such as future performance change trends, fluctuations, etc.).

[0132] In the embodiment of the present application, the performance characteristics and data characteristics are subjected to feature fusion processing to obtain fused data characteristics. Through feature fusion, the computer device can integrate the current state of the network and future change trends to obtain more comprehensive network performance characteristics.

[0133] In the above embodiment, by extracting features from network performance data and change trend data respectively, the key information in these two types of data can be fully mined, and the features of the two can be integrated to comprehensively consider the current network performance and future change trends, thereby providing a more accurate network path prediction.

[0134] In an exemplary embodiment, Figure 6 As shown, the above method may further include step 601 to step 602. Wherein:

[0135] Step 601, based on the fused data features, determine the joint loss function of the trend prediction model and the instruction generation model.

[0136] Among them, the joint loss function is a loss function used to simultaneously optimize the trend prediction model and the instruction generation model. It combines the losses of the trend prediction model and the instruction generation model into an overall loss. By minimizing the joint loss function, the trend prediction model and the instruction generation model can jointly improve performance.

[0137] In an embodiment of the present application, a joint loss function of a trend prediction model and an instruction generation model is determined based on fused data features.

[0138] In some embodiments, it is necessary to define a loss function that can simultaneously evaluate the performance of the trend prediction model and the instruction generation model. The loss function combines the goals of the trend prediction model and the instruction generation model: the goal of the trend prediction model is to accurately predict the changing trend of network performance, and the goal of the instruction generation model is to generate optimized network path switching instructions based on these predictions. The joint loss function weights and combines the errors of these two goals to form an overall loss metric. The joint loss function includes prediction errors (such as trend prediction errors) and instruction generation errors (such as path switching errors) to ensure that the trend prediction model and the instruction generation model can be optimized together during the training process to achieve the optimal network optimization effect.

[0139] In some embodiments, the loss value corresponding to the joint loss function can be obtained by first determining the evaluation rules corresponding to the trend prediction model and the instruction generation model, then obtaining the loss value corresponding to the evaluation rule based on the evaluation rule, and then weighting each loss value to obtain the final loss value. The evaluation rules may include:

[0140] (1) Network delay evaluation rules:

[0141] Define the maximum average network delay threshold as 50 milliseconds, calculate the average network delay over a period of time (such as 1 minute), and if the average network delay exceeds 50 milliseconds, the loss value = (actual average delay - 50) × 10.

[0142] (2) Bandwidth utilization evaluation rules:

[0143] Set the minimum bandwidth utilization target to 70% and calculate the actual bandwidth utilization. If the actual bandwidth utilization is lower than 70%, the loss value = (70-actual utilization) × 5.

[0144] (3) Network topology adaptability evaluation rules:

[0145] When the network topology changes, the timer starts and the analysis and decision-making module is required to respond within 10 seconds and adjust the network connection strategy. If there is no response within 10 seconds, the loss value = 20.

[0146] (4) Traffic burst processing evaluation rules:

[0147] Monitor the sudden increase of traffic, set the threshold to 50% increase in a short period of time (such as 30 seconds), and count the packet loss rate during the traffic burst. If the packet loss rate exceeds 5%, the loss value = (packet loss rate - 5) × 100.

[0148] (5) Safety assessment rules:

[0149] Check the data encryption status. If it is not encrypted, the loss value = 50. Monitor the number of attacks received. Every time the number of attacks exceeds the preset security attack threshold, the loss value = 10.

[0150] (6) Fault recovery time evaluation rules:

[0151] When a network failure occurs, the timer starts and the failure recovery time is required to be no more than 5 minutes. If it exceeds 5 minutes, the loss value = (actual recovery time - 5) × 5.

[0152] (7) Quality of Service (QoS) evaluation rules

[0153] Compare the availability guarantee indicators in the QoS agreement signed with the cloud service provider, such as the requirement of 99.9% availability, and calculate the actual availability. If the actual availability is lower than the agreement requirement, the loss value = (agreement availability - actual availability) × 1000.

[0154] Assume that the loss formula obtained by the above rules is:

[0155]

[0156] Among them, the loss value corresponding to the network delay evaluation rule is , the loss value corresponding to the bandwidth utilization evaluation rule is , the loss value corresponding to the network topology adaptability evaluation rule is , the loss value corresponding to the traffic burst processing evaluation rule is , the loss value corresponding to the safety assessment rule is , the loss value corresponding to the fault recovery time evaluation rule is , the loss value corresponding to the Quality of Service (QoS) evaluation rule is , and set the weight parameters respectively to , and finally add up to get the loss value, is a regularization term used to prevent overfitting.

[0157] Step 602, adjusting parameters of the trend prediction model and the instruction generation model based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0158] Among them, parameter adjustment refers to optimizing the performance of the trend prediction model and the instruction generation model by adjusting the model's hyperparameters (such as learning rate, regularization parameters, etc.) during the training process. In this process, the trend prediction model and the instruction generation model will be adjusted according to the feedback of the joint loss function to improve their performance in the task. Parameter adjustment usually uses methods such as gradient descent, grid search, random search, etc. to find the best hyperparameter combination to optimize the prediction accuracy of the trend prediction model and the instruction generation model and the effect of generating instructions.

[0159] In an embodiment of the present application, the computer device is based on the joint loss function, and the trend prediction model and the instruction generation model will jointly adjust parameters. Parameter adjustment processing includes optimization of model hyperparameters, such as learning rate, regularization coefficient, network structure, etc., to improve the accuracy and robustness of the model. In this process, the trend prediction model and the instruction generation model will adjust their parameters according to the feedback of the joint loss function to minimize errors and improve overall performance. For example, the trend prediction model may adjust its prediction accuracy to make the prediction of future network status more accurate; at the same time, the instruction generation model will generate more efficient path switching instructions based on these more accurate prediction results.

[0160] The adjusted trend prediction model and instruction generation model will be able to better predict the changing trends of network performance and generate more accurate network path switching instructions based on these predictions.

[0161] In the above embodiment, through the optimization of the joint loss function, the performance of the trend prediction model and the instruction generation model will be improved simultaneously, making the network path switching more intelligent and efficient, which helps to improve the overall stability and performance of the network. This can not only cope with the current state of the network, but also make forward-looking adjustments based on the predicted changing trends, further improving the utilization efficiency and responsiveness of network resources.

[0162] In an exemplary embodiment, the above-mentioned “executing a network path switching action based on a network path switching instruction so that the transmission path of the service data is switched from the current network path to the target network path” may include:

[0163] Based on the network path switching instruction, a label is allocated to the service data from a pre-configured label forwarding table, so that the service data is switched from the current network path to the target network path based on the label.

[0164] The label forwarding table is a mapping table used to store network path switching, which contains the mapping relationship between labels and network paths. During the network path switching process, labels are used to identify different network paths and corresponding traffic. The pre-configured label forwarding table saves these pre-defined label information so that when the path needs to be switched, the corresponding label can be quickly found and applied to forward the data flow.

[0165] A label is an identifier used to identify a specific data stream or network path during the network path switching process. By assigning a label to the service data, the transmission path is switched from the current network path to the target network path, and the correct routing and processing are performed based on the label.

[0166] In an embodiment of the present application, when performing path switching, the computer device needs to obtain a label corresponding to the target network path by searching a pre-configured label forwarding table. Then, based on the target network path label found in the label forwarding table, the computer device will assign a new label to the business data being transmitted. This label will indicate the new transmission path of the business data. When the business data is assigned a new label, the computer device will perform a data flow forwarding operation according to the label, switching the business data from the current network path to the target network path.

[0167] In the above embodiments, traditional network path switching requires a long time to recalculate the best path, or requires more complex signal exchange and protocol operations. However, with path switching based on label forwarding tables, after assigning labels, the computer device can quickly determine the data transmission path by checking the labels, thereby significantly reducing the delay of path switching.

[0168] In an exemplary embodiment, the method further includes:

[0169] The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

[0170] In the embodiment of the present application, after the path switching action is completed, the computer device needs to monitor the network performance data on the new path in real time and record it. The monitored network performance data will be stored and recorded to form a new network performance data set. The purpose of recording the network performance data set is to provide a basis for subsequent decision-making and network adjustment. In a dynamic environment, the network load and performance may change continuously, so it is necessary to continuously update these data to reflect the current network status.

[0171] The computer device can evaluate the effectiveness of the current network path based on the new network performance data set and adjust or optimize the network configuration as needed. For example, if the performance of a certain path is poor, the computer device may generate a new network path switching instruction to reroute the traffic to a path with better performance.

[0172] In the above embodiment, by continuously monitoring and recording the network performance after the path switching, the computer device can continuously adjust the network strategy according to the real-time data to improve the stability and performance of the overall network.

[0173] According to some embodiments of the present application, a network path switching method based on a multi-cloud network is provided, which is described by taking the method applied to a computer device as an example, and may include the following steps:

[0174] Step 1, perform iterative calculations on each sub-data set. During the first iteration, input the first sub-data set into the trend prediction model, use the trend prediction model to perform feature extraction on the first sub-data set, and obtain the change trend characteristics of the first iteration.

[0175] Step 2: During the non-first iteration process, the previous change trend characteristics obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model. The trend prediction model is used to perform feature extraction and feature fusion processing on the sub-data set of this iteration and the previous change trend characteristics in turn to obtain the change trend characteristics of this iteration.

[0176] Step 3: Determine the change trend characteristics obtained based on the last sub-data set as the change trend data.

[0177] Step 4: Use the instruction generation model to perform feature extraction processing on the network performance data set to obtain performance characteristics.

[0178] Step 5: Use the instruction generation model to perform feature extraction processing on the change trend data to obtain data features.

[0179] Step 6: Perform feature fusion processing on the performance features and data features to obtain fused data features.

[0180] Step 7: Adjust the fused data features based on the network optimization goal to obtain adjusted fused data features.

[0181] Step 8: Generate a network path switching instruction based on the fused data features and the adjusted fused data features.

[0182] In some embodiments, the trend prediction model and the instruction generation model may also be jointly trained, and the training process may include:

[0183] Based on the fusion data features, the joint loss function of the trend prediction model and the instruction generation model is determined. Based on the joint loss function, the trend prediction model and the instruction generation model are adjusted to optimize the trend prediction model and the instruction generation model.

[0184] Step 9: assign a label to the service data from a pre-configured label forwarding table based on the network path switching instruction, so that the service data is switched from the current network path to the target network path based on the label.

[0185] Step 10: Use the execution result corresponding to the network path switching action as a network performance data set for subsequent use.

[0186] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0187] Based on the same inventive concept, the embodiment of the present application also provides a network path switching device based on a multi-cloud network for implementing the network path switching method based on a multi-cloud network involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the network path switching device based on a multi-cloud network provided below can refer to the limitations of the network path switching method based on a multi-cloud network above, and will not be repeated here.

[0188] In an exemplary embodiment, Figure 7 As shown, a network path switching device based on a multi-cloud network is provided, including: a prediction processing module 701, an adjustment processing module 702 and a path switching module 703, wherein:

[0189] The prediction processing module 701 is used to perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0190] The adjustment processing module 702 is used to perform feature extraction processing and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction;

[0191] The path switching module 703 is used to execute a network path switching action based on the network path switching instruction, so that the transmission path of the service data is switched from the current network path to the target network path.

[0192] In an exemplary embodiment, the prediction processing module 701 is specifically used to perform iterative calculations on each sub-data set. During the first iteration, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend characteristics of the first iteration; during the non-first iteration, the previous change trend characteristics obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing and feature fusion processing on the sub-data set of this iteration and the previous change trend characteristics in turn to obtain the change trend characteristics of this iteration; the change trend characteristics obtained based on the last sub-data set are determined as the change trend data.

[0193] In an exemplary embodiment, the adjustment processing module 702 is specifically used to use the instruction generation model to perform feature extraction processing and feature fusion processing on the network performance data set and the change trend data respectively to obtain fused data features; adjust the fused data features based on the network optimization goals to obtain adjusted fused data features; generate network path switching instructions based on the fused data features and the adjusted fused data features.

[0194] In an exemplary embodiment, the above-mentioned adjustment processing module 702 is specifically used to use the instruction generation model to perform feature extraction processing on the network performance data set to obtain performance characteristics; use the instruction generation model to perform feature extraction processing on the change trend data to obtain data characteristics; perform feature fusion processing on the performance characteristics and data characteristics to obtain fused data characteristics.

[0195] In an exemplary embodiment, the above device may further include:

[0196] A loss function determination module 704 is used to determine a joint loss function of a trend prediction model and an instruction generation model based on fused data features;

[0197] The parameter adjustment module 705 is used to adjust the parameters of the trend prediction model and the instruction generation model based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0198] In an exemplary embodiment, the path switching module 703 is specifically configured to assign a label to the service data from a preconfigured label forwarding table based on the network path switching instruction, so that the service data is switched from the current network path to the target network path based on the label.

[0199] In an exemplary embodiment, the above device may further include:

[0200] Feedback module 705, used to use the execution result corresponding to the network path switching action as a network performance data set for subsequent use;

[0201] Each module in the above-mentioned network path switching device based on multi-cloud network can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0202] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0203] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0204] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0205] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0206] In one embodiment, the network performance data set includes a plurality of sub-data sets arranged in chronological order, and the processor further implements the following steps when executing the computer program:

[0207] Iterative calculation is performed for each sub-data set. In the first iteration process, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend features of the first iteration;

[0208] In the non-first iteration process, the last change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the sub-data set of this iteration and the last change trend feature are sequentially subjected to feature extraction and feature fusion processing by the trend prediction model to obtain the change trend feature of this iteration;

[0209] The change trend feature obtained based on the last sub-data set is determined as the change trend data.

[0210] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0211] The instruction generation model is used to perform feature extraction and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features;

[0212] Adjust the fused data features based on the network optimization target to obtain adjusted fused data features;

[0213] A network path switching instruction is generated based on the fused data features and the adjusted fused data features.

[0214] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0215] Using the instruction generation model to perform feature extraction processing on the network performance data set, the performance characteristics are obtained;

[0216] The instruction generation model is used to perform feature extraction processing on the change trend data to obtain data features;

[0217] The performance features and data features are fused to obtain fused data features.

[0218] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0219] Based on the fused data features, determine the joint loss function of the trend prediction model and the instruction generation model;

[0220] The trend prediction model and the instruction generation model are adjusted based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0221] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0222] Based on the network path switching instruction, a label is allocated to the service data from a pre-configured label forwarding table, so that the service data is switched from the current network path to the target network path based on the label.

[0223] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0224] The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

[0225] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0226] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0227] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0228] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0229] In one embodiment, the network performance data set includes a plurality of sub-data sets arranged in chronological order, and when the computer program is executed by the processor, the following steps are further implemented:

[0230] Iterative calculation is performed for each sub-data set. In the first iteration process, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend features of the first iteration;

[0231] In the non-first iteration process, the last change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the sub-data set of this iteration and the last change trend feature are sequentially subjected to feature extraction and feature fusion processing by the trend prediction model to obtain the change trend feature of this iteration;

[0232] The change trend feature obtained based on the last sub-data set is determined as the change trend data.

[0233] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0234] The instruction generation model is used to perform feature extraction and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features;

[0235] Adjust the fused data features based on the network optimization target to obtain adjusted fused data features;

[0236] A network path switching instruction is generated based on the fused data features and the adjusted fused data features.

[0237] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0238] Using the instruction generation model to perform feature extraction processing on the network performance data set, the performance characteristics are obtained;

[0239] The instruction generation model is used to perform feature extraction processing on the change trend data to obtain data features;

[0240] The performance features and data features are fused to obtain fused data features.

[0241] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0242] Based on the fused data features, determine the joint loss function of the trend prediction model and the instruction generation model;

[0243] The trend prediction model and the instruction generation model are adjusted based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0244] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0245] Based on the network path switching instruction, a label is allocated to the service data from a pre-configured label forwarding table, so that the service data is switched from the current network path to the target network path based on the label.

[0246] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0247] The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

[0248] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0249] Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data;

[0250] Perform feature extraction and adjustment processing on the network performance data set and change trend data in sequence to obtain a network path switching instruction;

[0251] A network path switching action is executed based on the network path switching instruction, so that the transmission path of the business data is switched from the current network path to the target network path.

[0252] In one embodiment, the network performance data set includes a plurality of sub-data sets arranged in chronological order, and when the computer program is executed by the processor, the following steps are further implemented:

[0253] Iterative calculation is performed for each sub-data set. In the first iteration process, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend features of the first iteration;

[0254] In the non-first iteration process, the last change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the sub-data set of this iteration and the last change trend feature are sequentially subjected to feature extraction and feature fusion processing by the trend prediction model to obtain the change trend feature of this iteration;

[0255] The change trend feature obtained based on the last sub-data set is determined as the change trend data.

[0256] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0257] The instruction generation model is used to perform feature extraction and feature fusion processing on the network performance data set and the change trend data, respectively, to obtain fused data features;

[0258] Adjust the fused data features based on the network optimization target to obtain adjusted fused data features;

[0259] A network path switching instruction is generated based on the fused data features and the adjusted fused data features.

[0260] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0261] Using the instruction generation model to perform feature extraction processing on the network performance data set, the performance characteristics are obtained;

[0262] The instruction generation model is used to perform feature extraction processing on the change trend data to obtain data features;

[0263] The performance features and data features are fused to obtain fused data features.

[0264] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0265] Based on the fused data features, determine the joint loss function of the trend prediction model and the instruction generation model;

[0266] The trend prediction model and the instruction generation model are adjusted based on the joint loss function to optimize the trend prediction model and the instruction generation model.

[0267] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0268] Based on the network path switching instruction, a label is allocated to the service data from a pre-configured label forwarding table, so that the service data is switched from the current network path to the target network path based on the label.

[0269] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0270] The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

[0271] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0272] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0273] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A network path switching method based on a multi-cloud network, characterized in that: The method comprises: Perform trend prediction processing based on the network performance data set of each network path acquired in advance to obtain change trend data; Performing feature extraction processing and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction; A network path switching action is executed based on the network path switching instruction, so that the transmission path of the service data is switched from the current network path to the target network path.

2. The method according to claim 1, characterized in that The network performance data set includes a plurality of sub-data sets arranged in chronological order, and the network performance data set based on each network path acquired in advance is subjected to trend prediction processing to obtain change trend data, including: Iterative calculation is performed for each of the sub-data sets. In the first iteration process, the first sub-data set is input into the trend prediction model, and the trend prediction model is used to perform feature extraction processing on the first sub-data set to obtain the change trend feature of the first iteration; In a non-first iteration process, the last change trend feature obtained by the trend prediction model and the sub-data set of this iteration are input into the trend prediction model, and the sub-data set of this iteration and the last change trend feature are sequentially subjected to feature extraction processing and feature fusion processing by using the trend prediction model to obtain the change trend feature of this iteration; The change trend feature obtained based on the last sub-data set is determined as the change trend data.

3. The method according to claim 2, characterized in that The step of sequentially performing feature extraction processing and adjustment processing on the network performance data set and the change trend data to obtain a network path switching instruction includes: Using the instruction generation model, the network performance data set and the change trend data are subjected to feature extraction and feature fusion processing respectively to obtain fused data features; Adjusting the fused data features based on the network optimization target to obtain adjusted fused data features; The network path switching instruction is generated based on the fused data feature and the adjusted fused data feature.

4. The method according to claim 3, characterized in that The using instruction generation model to perform feature extraction processing and feature fusion processing on the network performance data set and the change trend data respectively to obtain fused data features includes: Using the instruction generation model to perform feature extraction processing on the network performance data set to obtain performance features; Using the instruction generation model to perform feature extraction processing on the change trend data to obtain data features; The performance feature and the data feature are subjected to feature fusion processing to obtain the fused data feature.

5. The method according to claim 3, characterized in that: The method further comprises: Determining a joint loss function of the trend prediction model and the instruction generation model based on the fused data features; The trend prediction model and the instruction generation model are adjusted based on the joint loss function to optimize the trend prediction model and the instruction generation model.

6. The method according to claim 1, characterized in that The performing the network path switching action based on the network path switching instruction so as to switch the transmission path of the service data from the current network path to the target network path includes: Based on the network path switching instruction, a label is allocated to the service data from a preconfigured label forwarding table, so that the service data is switched from the current network path to the target network path based on the label.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: The execution result corresponding to the network path switching action is used as a network performance data set for subsequent use.

8. A network path switching device based on a multi-cloud network, characterized in that: The device comprises: A prediction processing module is used to perform trend prediction processing based on a set of network performance data of each network path acquired in advance to obtain change trend data; An adjustment processing module, used to perform feature extraction processing and adjustment processing on the network performance data set and the change trend data in sequence to obtain a network path switching instruction; The path switching module is used to execute a network path switching action based on the network path switching instruction so that the transmission path of the service data is switched from the current network path to the target network path.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images