Traffic scheduling methods, apparatus, electronic devices and storage media

Abstract

This application relates to the field of network technology, and provides a traffic scheduling method, apparatus, electronic device, and storage medium. The method includes: acquiring historical CDN traffic data, switch historical traffic data, and metropolitan area network (MAN) traffic data of network nodes; inputting the historical CDN traffic data, switch historical traffic data, and MAN historical traffic data into a traffic prediction model to obtain the predicted CDN traffic, switch predicted traffic, and MAN predicted traffic for future moments output by the traffic prediction model; determining the inbound traffic and outbound traffic of network nodes based on the predicted CDN traffic, switch predicted traffic, and MAN predicted traffic; and performing traffic scheduling based on the inbound traffic and outbound traffic. This method effectively avoids scheduling lag, achieves high scheduling accuracy, and, by using a model for traffic prediction, eliminates reliance on the experience of scheduling personnel, improving scheduling efficiency and reducing the cost of manual operation and maintenance.

Description

Technical Field

[0001] This application relates to the field of network technology, specifically to a traffic scheduling method, apparatus, electronic device, and storage medium. Background Technology

[0002] To improve the traffic utilization of each network node in the Content Delivery Network (CDN) of a telecommunications operator's broadband TV service, it is necessary to schedule the traffic of each network node through a reasonable scheduling method.

[0003] Currently, most traffic scheduling is carried out using static scheduling strategies configured manually. This scheduling strategy is lagging, easily affected by other factors, and relies heavily on the relevant experience of the scheduling personnel. The strategy configuration method is also cumbersome. Summary of the Invention

[0004] This application provides a traffic scheduling method, apparatus, electronic device, and storage medium to solve the technical problem that prior art scheduling strategies are lagging and heavily reliant on the relevant experience of scheduling personnel.

[0005] In a first aspect, embodiments of this application provide a traffic scheduling method, comprising: acquiring historical CDN traffic data, historical switch traffic data, and historical metropolitan area network (MAN) traffic data of network nodes; inputting the historical CDN traffic data, historical switch traffic data, and historical MAN traffic data into a traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic at future times output by the traffic prediction model; determining the inbound traffic and outbound traffic of network nodes based on the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic; and performing traffic scheduling based on the inbound traffic and outbound traffic; wherein the traffic prediction model is trained based on sample historical CDN traffic data and sample CDN predicted traffic corresponding to the sample historical CDN traffic data, sample historical switch traffic data and sample predicted switch traffic corresponding to the sample historical switch traffic data, and sample historical MAN traffic data and sample predicted MAN traffic corresponding to the sample historical MAN traffic data.

[0006] In one embodiment, the traffic prediction model includes a decomposition network and an additive network. The decomposition network decomposes CDN historical traffic data, switch historical traffic data, and metropolitan area network (MAN) historical traffic data to generate trend components, periodic components, holiday components, and error components. The additive network fits and predicts the trend components, periodic components, holiday components, and error components to generate predicted CDN traffic, switch traffic, and MAN traffic for future time periods. The trend components characterize the non-periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components characterize the periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components are represented by Fourier series. The holiday components characterize the impact of non-periodic holidays on CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The error components characterize unpredictable fluctuations in the model.

[0007] In one embodiment, inputting historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data into a traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted metropolitan area network traffic for future moments output by the traffic prediction model includes: inputting the historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data into the decomposition network of the traffic prediction model to obtain the trend component, periodic component, holiday component, and error component output by the decomposition network; and inputting the trend component, periodic component, holiday component, and error component into the addition network of the traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted metropolitan area network traffic for future moments output by the addition network.

[0008] In one embodiment, before determining the inbound and outbound traffic of a network node based on CDN predicted traffic, switch predicted traffic, and metropolitan area network (MAN) predicted traffic, the method further includes: determining data mutation compensation for CDN predicted traffic, data mutation compensation for switch predicted traffic, and data mutation compensation for MAN predicted traffic; and performing superposition calculations on the data mutation compensation and CDN predicted traffic, the data mutation compensation and switch predicted traffic, and the data mutation compensation and MAN predicted traffic of MAN predicted traffic to obtain the CDN predicted traffic, switch predicted traffic, and MAN predicted traffic at future moments after compensation processing.

[0009] In one embodiment, determining the available inbound traffic and the required outbound traffic of a network node based on CDN predicted traffic, switch predicted traffic, and metropolitan area network (MAN) predicted traffic includes: determining whether a network node meets scheduling conditions based on CDN predicted traffic and switch predicted traffic; if the network node meets the scheduling conditions, determining CDN inbound traffic and CDN outbound traffic based on CDN predicted traffic and CDN traffic safety line, determining switch inbound traffic and switch outbound traffic based on switch predicted traffic and switch traffic safety line, and determining MAN inbound traffic and MAN outbound traffic based on MAN predicted traffic and MAN traffic maximum value; determining the available inbound traffic of a network node based on CDN inbound traffic, switch inbound traffic, and MAN inbound traffic; and determining the required outbound traffic of a network node based on CDN outbound traffic, switch outbound traffic, and MAN outbound traffic.

[0010] In one embodiment, the number of network nodes is at least two; traffic scheduling based on inbound traffic and outbound traffic includes: classifying network nodes based on the inbound traffic and outbound traffic of each network node to obtain at least one traffic inbound node and at least one traffic outbound node; determining the traffic inbound node corresponding to each traffic outbound node based on the outbound traffic of each traffic outbound node and the traffic utilization rate of each traffic inbound node; and scheduling the outbound traffic of each traffic outbound node to the corresponding traffic inbound node.

[0011] In one embodiment, after traffic scheduling is performed based on inbound and outbound traffic, the method further includes: calculating scheduling compensation for CDN predicted traffic, scheduling compensation for switch predicted traffic, and scheduling compensation for metropolitan area network predicted traffic based on the least squares method; wherein, the scheduling compensation for CDN predicted traffic is used to compensate for the CDN predicted traffic output by the traffic prediction model in the next prediction period; the scheduling compensation for switch predicted traffic is used to compensate for the switch predicted traffic output by the traffic prediction model in the next prediction period; and the scheduling compensation for metropolitan area network predicted traffic is used to compensate for the metropolitan area network predicted traffic output by the traffic prediction model in the next prediction period.

[0012] Secondly, embodiments of this application provide a traffic scheduling device, comprising: an acquisition module for acquiring historical CDN traffic data, historical switch traffic data, and historical metropolitan area network (MAN) traffic data of network nodes; a prediction module for inputting the historical CDN traffic data, historical switch traffic data, and historical MAN traffic data into a traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic at future times output by the traffic prediction model; a determination module for determining the inbound traffic and outbound traffic of network nodes based on the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic; and a scheduling module for performing traffic scheduling based on the inbound traffic and outbound traffic; wherein the traffic prediction model is trained based on sample historical CDN traffic data and sample CDN predicted traffic corresponding to the sample historical CDN traffic data, sample historical switch traffic data and sample predicted switch traffic corresponding to the sample historical switch traffic data, and sample historical MAN traffic data and sample predicted MAN traffic corresponding to the sample historical MAN traffic data.

[0013] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the traffic scheduling methods described above.

[0014] Fourthly, embodiments of this application provide a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the traffic scheduling method as described above.

[0015] The traffic scheduling method, apparatus, electronic device, and storage medium provided in this application acquire historical CDN traffic data, historical switch traffic data, and historical metropolitan area network (MAN) traffic data of network nodes; input the historical CDN traffic data, historical switch traffic data, and historical MAN traffic data into a traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic for future moments output by the traffic prediction model; determine the inbound traffic and outbound traffic of network nodes based on the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic; and perform traffic scheduling based on the inbound traffic and outbound traffic. The traffic prediction model is trained based on sample historical CDN traffic data and corresponding sample CDN predicted traffic, sample historical switch traffic data and corresponding sample predicted switch traffic, and sample historical MAN traffic data and corresponding sample predicted MAN traffic. By using the above method, historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data of network nodes are input into the traffic prediction model. The traffic prediction model predicts the CDN traffic, switch traffic, and metropolitan area network traffic of network nodes in advance. Based on the prediction results, traffic scheduling of network nodes can be performed in advance, which can effectively avoid scheduling lag, achieve high scheduling accuracy, and eliminate the reliance on the relevant experience of scheduling personnel through model-based traffic prediction, thereby improving scheduling efficiency and reducing the cost of manual operation and maintenance. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the traffic scheduling method provided in the embodiments of this application;

[0018] Figure 2 This is a second flowchart illustrating the traffic scheduling method provided in the embodiments of this application;

[0019] Figure 3 This is a flowchart illustrating the intelligent scheduling method provided in an embodiment of this application;

[0020] Figure 4 This is a flowchart illustrating the process of determining the maximum value of the future outgoing traffic, as provided in an embodiment of this application.

[0021] Figure 5This is a schematic diagram of the traffic scheduling device provided in the embodiments of this application;

[0022] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Please see Figure 1 , Figure 1 This is one of the flowcharts illustrating the traffic scheduling method provided in this application embodiment. In this embodiment, the traffic scheduling method includes steps S110 to S140, each step being as follows:

[0025] S110: Obtain historical CDN traffic data, switch historical traffic data, and metropolitan area network traffic data of network nodes.

[0026] Please see Figure 2 , Figure 2 This is the second flowchart of the traffic scheduling method provided in the embodiments of this application.

[0027] Specifically, such as Figure 2 As shown, the historical traffic data of CDN, switches, and metropolitan area network for each network node can be directly read from the database.

[0028] S120: Input CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data into the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future moments output by the traffic prediction model.

[0029] The traffic prediction model is trained based on sample CDN historical traffic data and corresponding sample CDN predicted traffic, sample switch historical traffic data and corresponding sample switch predicted traffic, and sample metropolitan area network historical traffic data and corresponding sample metropolitan area network predicted traffic.

[0030] S130: Based on CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic, determine the inbound traffic and outbound traffic of network nodes.

[0031] Generally, there are multiple network nodes, and the incoming traffic and outgoing traffic of each network node are different. Therefore, the incoming traffic and outgoing traffic of each network node can be determined based on the CDN predicted traffic, switch predicted traffic and metropolitan area network predicted traffic of each network node at future time points obtained by the traffic prediction model.

[0032] S140: Traffic scheduling is performed based on incoming traffic and outgoing traffic.

[0033] Specifically, traffic scheduling is performed on all network nodes based on the incoming and outgoing traffic of each network node.

[0034] The traffic scheduling method provided in this application involves acquiring historical CDN traffic data, historical switch traffic data, and historical metropolitan area network (MAN) traffic data of network nodes; inputting these historical traffic data into a traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic for future moments output by the traffic prediction model; determining the inbound and outbound traffic of network nodes based on the predicted CDN traffic, predicted switch traffic, and predicted MAN traffic; and performing traffic scheduling based on the inbound and outbound traffic. The traffic prediction model is trained based on sample historical CDN traffic data and corresponding sample CDN predicted traffic, sample historical switch traffic data and corresponding sample predicted switch traffic, and sample historical MAN traffic data and corresponding sample predicted MAN traffic. By using the above method, historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data of network nodes are input into the traffic prediction model. The traffic prediction model predicts the CDN traffic, switch traffic, and metropolitan area network traffic of network nodes in advance. Based on the prediction results, traffic scheduling of network nodes can be performed in advance, which can effectively avoid scheduling lag, achieve high scheduling accuracy, and eliminate the reliance on the relevant experience of scheduling personnel through model-based traffic prediction, thereby improving scheduling efficiency and reducing the cost of manual operation and maintenance.

[0035] In some embodiments, the traffic prediction model includes a decomposition network and an additive network. The decomposition network decomposes CDN historical traffic data, switch historical traffic data, and metropolitan area network (MAN) historical traffic data to generate trend components, periodic components, holiday components, and error components. The additive network fits and predicts the trend components, periodic components, holiday components, and error components to generate predicted CDN traffic, switch traffic, and MAN traffic for future time periods. The trend components characterize the non-periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components characterize the periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components are represented by Fourier series. The holiday components characterize the impact of non-periodic holidays on CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The error components characterize unpredictable fluctuations in the model.

[0036] In this embodiment, the traffic prediction model is an improved Prophet model.

[0037] The Prophet model is a time series model based on the concepts of additive modeling and decomposition, incorporating multiple components such as trend, seasonality, and holiday effects. Specifically, the Prophet model can decompose input data (time series) into three components: trend, seasonality, and holidays, and then model and predict each component separately.

[0038] Specifically, the traffic prediction model includes decomposition networks and additive networks.

[0039] The decomposition network is used to decompose CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data to generate trend component g(t), periodic component s(t), holiday component h(t), and error component ε. t .

[0040] Among them, the trend component g(t) is used to characterize the non-periodic change trend of CDN historical traffic data, switch historical traffic data and metropolitan area network historical traffic data.

[0041] The periodic component s(t) is used to characterize the periodic change trend of CDN historical traffic data, switch historical traffic data and metropolitan area network historical traffic data; the periodic component s(t) is also called the seasonal component, which is measured in weeks or years.

[0042] The periodic component s(t) is represented by a Fourier series.

[0043] The holiday component h(t) is used to characterize the impact of non-periodic holidays on CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data.

[0044] Error component ε t The error term, also known as the residual term, is used to characterize unpredictable fluctuations in the model.

[0045] Error component ε t It follows a Gaussian distribution.

[0046] Additive networks are used to process the trend component g(t), periodic component s(t), holiday component h(t), and error component ε. t Perform fitting predictions to generate predicted CDN traffic, switch traffic, and metropolitan area network traffic for future time periods.

[0047] Therefore, the specific expression for the traffic prediction model is as follows:

[0048] y(t)=g(t)+s(t)+h(t)+∈ t ;

[0049] Where y(t) represents the output of the traffic prediction model.

[0050] Alternatively, the trend component g(t) can be represented using a piecewise linear function, and the specific expression for the trend component g(t) is as follows:

[0051] g(t)=(k+a(t) T δ)t+(m+a(t) T γ);

[0052] Where k is the growth rate; δ is the change in the growth rate; m is the offset; γ is the change in the offset; a(t) T t represents the model parameters; t is the independent variable representing the change over time.

[0053] Optionally, the trend component g(t) can be determined through logistic regression modeling, and the specific expression for the trend component g(t) is as follows:

[0054]

[0055] Where k is the growth rate; δ is the change in the growth rate; m is the offset; γ is the change in the offset; a(t) T C(t) are model parameters; exp represents an exponential function with the natural constant e as the base.

[0056] Optionally, the periodic component s(t) is modeled using a Fourier series. The specific expression for the periodic component s(t) in Fourier series form is as follows:

[0057]

[0058] Among them, a n b n is the Fourier coefficient; N is the number of terms, which can be adjusted to change the flexibility of the model; P is the period, if the input data is annual data then P = 365.25, if the input data is weekly data then P = 7.

[0059] Optionally, the holiday component h(t) can be represented by a simple indicator function, which can assign an additional value on the holiday day. The specific expression for the holiday component h(t) is as follows:

[0060] h(t)=Z(t)κ;

[0061] Z(t) = [1(t∈D1),...,1(t∈D1)] L )];

[0062] Where Di is the date of the i-th holiday of each year; κ is the prior distribution added to this item. For each holiday, Di needs to specify a parameter κ to represent the impact caused by this holiday; Z(t) is the regression matrix.

[0063] Generally, in addition to daily and weekly cycles, CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data will show significant changes during special holidays, such as Lunar New Year's Eve and National Day.

[0064] However, in the existing Prophet model, the holiday item (i.e., the holiday component) is implemented using an indicator function on a daily basis. This results in the holiday item having the same value every day, but the actual impact of major holidays and events is concentrated within a certain period of time. Therefore, the holiday item in the Prophet model needs to be adjusted.

[0065] Preferably, a Fourier series with 10 terms is used to represent the impact of holidays, i.e.:

[0066]

[0067] The period T is set to 365.25, and date conditions are set so that the effects of holidays only occur on specific dates, thereby improving the modeling effect of holiday items.

[0068] In some embodiments, inputting CDN historical traffic data, switch historical traffic data, and metropolitan area network (MAN) historical traffic data into a traffic prediction model to obtain the future predicted CDN traffic, switch traffic, and MAN traffic output by the traffic prediction model includes: inputting the CDN historical traffic data, switch historical traffic data, and MAN historical traffic data into the decomposition network of the traffic prediction model to obtain the trend component, periodic component, holiday component, and error component output by the decomposition network; and inputting the trend component, periodic component, holiday component, and error component into the addition network of the traffic prediction model to obtain the future predicted CDN traffic, switch traffic, and MAN traffic output by the addition network.

[0069] Please continue reading. Figure 2 The traffic prediction model is deployed on the server side. After obtaining historical traffic data of CDN, switches and metropolitan area networks from network nodes, the model prediction module needs to send the obtained historical data to the server for processing.

[0070] The server side includes a data preprocessing module, a model prediction module, a data mutation error compensation module, and an intelligent scheduling module.

[0071] Understandably, before inputting the acquired historical data into the model prediction module, it is necessary to input the acquired historical data into the data preprocessing module for data preprocessing.

[0072] The Prophet model is a univariate prediction model, and it specifies that the input data variables must be represented as "ds" and "y". Therefore, the data preprocessing module needs to organize the CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data into a data frame, with one column for each variable and one column for time. Each column of the data frame should contain the historical data of one variable and the name of that variable; the time column is named "ds", and the name of each of the other variable columns is "y".

[0073] After the data preprocessing module completes the data preprocessing, it inputs the obtained historical data into the traffic prediction model of the model prediction module.

[0074] Specifically, historical traffic data from CDN, switches, and metropolitan area networks are input into the decomposition network of the traffic prediction model to obtain the trend component, periodic component, holiday component, and error component output by the decomposition network. The trend component, periodic component, holiday component, and error component are then input into the addition network of the traffic prediction model to obtain the predicted CDN traffic, predicted switch traffic, and predicted metropolitan area network traffic at future moments output by the addition network.

[0075] The traffic scheduling method provided in this application can predict the possible changes of network nodes in the future based on historical data, and thus can schedule in advance for future changes at the current time, avoiding scheduling delays that lead to a decline in user experience and achieving high scheduling accuracy.

[0076] In some embodiments, before determining the inbound and outbound traffic of a network node based on CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic, the method further includes: determining data mutation compensation for CDN predicted traffic, data mutation compensation for switch predicted traffic, and data mutation compensation for metropolitan area network predicted traffic; and performing superposition calculations on the data mutation compensation and CDN predicted traffic, the data mutation compensation and switch predicted traffic, and the data mutation compensation and metropolitan area network predicted traffic, respectively, to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future moments after compensation processing.

[0077] The Prophet model can achieve good prediction results for ordinary data. However, when there are sudden changes in historical data, due to the global modeling characteristics of the Prophet model, all historical data will be given the same weight in the prediction process. Therefore, changes in local data may not have a significant impact on the overall prediction of the model.

[0078] For data far removed from the prediction time, this might only represent a temporary fluctuation that will subsequently return to normal, making a normal prediction result reasonable. However, if this situation occurs closer to the prediction time, maintaining the original prediction method will not accurately reflect the changes in the current data. To address this, compensation is needed for data abrupt changes closer to the prediction time to improve the accuracy of the model's predictions.

[0079] To address the aforementioned improvement needs, compensation can be calculated by modeling the mutation data. The calculated compensation can then be superimposed with the model's prediction results to improve the accuracy of the prediction results.

[0080] Specifically, data mutation compensation for CDN predicted traffic, data mutation compensation for switch predicted traffic, and data mutation compensation for metropolitan area network predicted traffic are determined.

[0081] Optionally, the data mutation compensation for CDN predicted traffic, the data mutation compensation for switch predicted traffic, and the data mutation compensation for metropolitan area network predicted traffic are determined based on the mobile weighted average method.

[0082] Specifically, the reference weights for data mutations are different. For the prediction time, the closer the mutation data is to the prediction time, the stronger its reference value, and the greater its weight should be. Therefore, the moving weighted average method is used to calculate the data mutation compensation. The compensation calculation formula for the data mutation compensation yt within this prediction period is as follows:

[0083] yt = (1-α)yt-1 + αxt;

[0084] Where x is the error value; α is the moving weighted average parameter; t is the independent variable representing time variation; and yt-1 is the compensation for data mutations in the previous prediction period.

[0085] Furthermore, after calculating the data mutation compensation for CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic using the compensation calculation formula, the data mutation compensation for CDN predicted traffic and CDN predicted traffic are superimposed (i.e., the data mutation compensation for CDN predicted traffic is superimposed as a fixed value onto CDN predicted traffic) to obtain the CDN predicted traffic at future times after compensation processing; the data mutation compensation for switch predicted traffic and switch predicted traffic are superimposed to obtain the switch predicted traffic at future times after compensation processing; the data mutation compensation for metropolitan area network predicted traffic and metropolitan area network predicted traffic are superimposed to obtain the metropolitan area network predicted traffic at future times after compensation processing.

[0086] Optionally, data mutation compensation for CDN predicted traffic, data mutation compensation for switch predicted traffic, and data mutation compensation for metropolitan area network predicted traffic are determined based on the Prophet model.

[0087] Specifically, the difference between the traffic data previously predicted by the Prophet model and the actual traffic data is plotted to obtain a traffic curve that changes over time. The data from the traffic curve is then input into a new Prophet model. The Prophet model learns the changes in the difference between the predicted traffic data and the actual traffic data to generate data mutation compensation for CDN predicted traffic, data mutation compensation for switch predicted traffic, and data mutation compensation for metropolitan area network predicted traffic.

[0088] Furthermore, the data mutation compensation and CDN predicted traffic, the data mutation compensation and switch predicted traffic, and the data mutation compensation and metropolitan area network predicted traffic are superimposed and calculated to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times after compensation processing.

[0089] like Figure 2As shown, after obtaining the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic, the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic are input into the data mutation error compensation module for compensation processing of the prediction results.

[0090] Generally, both methods can achieve good compensation, thereby improving the prediction effect. Therefore, different methods can be adopted according to different needs and usage scenarios.

[0091] The moving weighted average compensation method is mainly implemented through mathematical calculations. It has high execution efficiency and requires less sample data, but its accuracy is slightly lower. Therefore, it is mainly used in scenarios with a small number of samples or high operational efficiency requirements.

[0092] Model learning compensation requires using a new model to find the optimal compensation. It has high accuracy, but requires a large amount of sample data and computation. It is mainly used in scenarios where the number of samples is sufficient and higher accuracy is required.

[0093] The traffic scheduling method provided in this application compensates the model prediction results by calculating the data mutation compensation, which improves the problem that a large amount of data prediction cannot effectively perceive short-term large fluctuations, and can effectively improve the accuracy of the prediction results.

[0094] In some embodiments, determining the available inbound traffic and the required outbound traffic of a network node based on CDN predicted traffic, switch predicted traffic, and metropolitan area network (MAN) predicted traffic includes: determining whether a network node meets scheduling conditions based on CDN predicted traffic and switch predicted traffic; if a network node meets scheduling conditions, determining CDN inbound traffic and CDN outbound traffic based on CDN predicted traffic and CDN traffic safety line, determining switch inbound traffic and switch outbound traffic based on switch predicted traffic and switch traffic safety line, and determining MAN inbound traffic and MAN outbound traffic based on MAN predicted traffic and MAN traffic maximum value; determining the available inbound traffic of a network node based on CDN inbound traffic, switch inbound traffic, and MAN inbound traffic; and determining the required outbound traffic of a network node based on CDN outbound traffic, switch outbound traffic, and MAN outbound traffic.

[0095] After supplementing the prediction results, the intelligent scheduling module can perform intelligent scheduling based on the prediction results.

[0096] Please see Figure 3 , Figure 3 This is a flowchart illustrating the intelligent scheduling method provided in the embodiments of this application.

[0097] Specifically, based on the predicted data, namely CDN predicted traffic and switch predicted traffic, it is determined whether the network node meets the scheduling conditions.

[0098] Each network node is equipped with a traffic safety line and a traffic scheduling line. If a network node meets the scheduling conditions, traffic scheduling is required.

[0099] In this embodiment, the scheduling conditions are as follows:

[0100] (1) The CDN predicted traffic is greater than or equal to the CDN traffic scheduling line;

[0101] (2) The predicted traffic of the switch is greater than or equal to the traffic scheduling line of the switch.

[0102] A network node can be considered to meet the scheduling conditions if either the CDN predicted traffic is greater than or equal to the CDN traffic scheduling line or the switch predicted traffic is greater than or equal to the switch traffic scheduling line.

[0103] Furthermore, if network nodes meet the scheduling conditions, CDN incoming and CDN outgoing traffic are determined based on CDN predicted traffic and CDN traffic safety line; switch incoming and switch outgoing traffic are determined based on switch predicted traffic and switch traffic safety line; and metropolitan area network incoming and metropolitan area network outgoing traffic are determined based on metropolitan area network predicted traffic and metropolitan area network maximum traffic value.

[0104] If a network node meets the scheduling conditions, then the traffic needs to be reduced to the traffic safety threshold.

[0105] Specifically, CDN outbound traffic is the difference between CDN predicted traffic and CDN traffic safety line; switch outbound traffic is the difference between switch predicted traffic and switch traffic safety line; metropolitan area network (MAN) predicted traffic includes MAN predicted incoming traffic and MAN predicted outbound traffic. MAN outbound traffic (i.e., the maximum MAN outbound traffic) is the difference between the maximum MAN traffic value and the MAN predicted incoming traffic.

[0106] Similarly, CDN inbound traffic is the difference between CDN predicted traffic and CDN traffic safety line; switch inbound traffic is the difference between switch predicted traffic and switch traffic safety line; metropolitan area network (MAN) inbound traffic (i.e., the maximum MAN inbound traffic) is the difference between the maximum MAN traffic value and the MAN predicted outbound traffic.

[0107] Furthermore, based on CDN inbound traffic, switch inbound traffic, and metropolitan area network inbound traffic, the available inbound traffic Vin of a network node is determined. The formula for calculating the available inbound traffic Vin of a network node is as follows:

[0108] Vin=Min(Vcdnin, Vswitchin, Vcmnetin);

[0109] Wherein, Min is the minimum value function; Vcdnin is the CDN incoming traffic; Vswitchin is the switch incoming traffic; and Vcmnetin is the metropolitan area network incoming traffic (i.e., the maximum incoming traffic of the metropolitan area network).

[0110] Furthermore, based on CDN outbound traffic, switch outbound traffic, and metropolitan area network outbound traffic, the required outbound traffic Vout of the network node is determined. The calculation formula for the required outbound traffic Vout of the network node is as follows:

[0111] Vout=Min(Max(Vcdnout, Vswitchout), Vcmnetout);

[0112] Where Max is the maximum value function; Vcdnout is the CDN outbound traffic; Vswitchout is the switch outbound traffic; and Vcmnetout is the metropolitan area network outbound traffic (i.e., the maximum outbound traffic of the metropolitan area network).

[0113] Since advance scheduling is based on future time-to-time conditions and is performed at the current time, a scheduling method needs to be designed to ensure that the current time-to-time scheduling meets the future time-to-in and future-to-out requirements.

[0114] For the traffic that needs to be dispatched by each network node, the traffic that needs to be dispatched at the current moment can be obtained according to the ratio. That is, the maximum value of the traffic that each network node needs to dispatch in the future within the preset scheduling time is determined. Based on the ratio of the change in traffic utilization rate of each network node at the current moment and the traffic utilization rate at the moment corresponding to the maximum value of the traffic that needs to be dispatched in the future, the traffic that needs to be dispatched at the current moment in order to meet the traffic that needs to be dispatched in the future can be calculated.

[0115] Please see Figure 4 , Figure 4 This is a flowchart illustrating the process of determining the maximum value of future outgoing traffic, as provided in an embodiment of this application.

[0116] For the predicted outbound traffic of each network node at the predicted future time, an iterative peak query is performed to determine the maximum value of the future outbound traffic. A basic scheduling duration T0 is set. First, it is determined whether scheduling is needed within the basic scheduling duration T0 from the start of the prediction. If scheduling is needed, the maximum value of the future outbound traffic of each network node is determined within this basic scheduling duration T0, and scheduling is performed based on this maximum value, thereby reducing the scheduling frequency and completing the scheduling in one step. If it is found that the maximum value of the future outbound traffic of each network node continues to increase within the basic scheduling duration T0, a buffer scheduling duration is added to the basic scheduling duration T0, and the search for the maximum value of the future outbound traffic continues until the maximum value of the future outbound traffic no longer changes or the basic scheduling duration plus the buffer scheduling duration has reached the total prediction duration T.

[0117] Generally, the higher the traffic utilization rate of a network node, the greater the traffic needs to be transferred out. Therefore, the maximum value of the traffic to be transferred out can be determined based on the traffic utilization rate of the network node.

[0118] Specifically, such as Figure 4 As shown, a basic scheduling duration T0 is set, and it is first determined whether scheduling needs to be carried out within the basic scheduling duration T0 from the prediction start time.

[0119] If scheduling is required, the traffic utilization rate A of the network nodes within the basic scheduling duration T0 is first obtained. If it is found that the traffic utilization rate A of each network node is still increasing within the basic scheduling duration T0, a buffer scheduling duration is added on the basis of the basic scheduling duration T0, and the traffic utilization rate B of the network nodes within T0+buffer scheduling duration is obtained.

[0120] Furthermore, determine whether the traffic utilization rate A is equal to the traffic utilization rate B, or determine whether the basic scheduling duration T0 + buffer scheduling duration reaches the predicted total duration T.

[0121] If traffic utilization rate A is equal to traffic utilization rate B, it means that the traffic utilization rate within the basic scheduling duration + buffer scheduling duration will no longer change. That is, when the traffic utilization rate is A, the traffic that the network node needs to send out is the maximum value of the traffic that the network node needs to send out in the future.

[0122] If the basic scheduling duration T0 + buffer scheduling duration does not reach the predicted total duration T, then when the traffic utilization rate is A, the traffic that the network node needs to send out is the maximum value of the traffic that the network node needs to send out in the future.

[0123] If the flow utilization rate A is not equal to the flow utilization rate B, and the basic scheduling duration T0 + buffer scheduling duration has not reached the predicted total duration T, then the flow utilization rate A is updated, that is, the value of the flow utilization rate B is assigned to A, denoted as A = B (where “=" means assignment), and the buffer scheduling duration Ta is added to the current basic scheduling duration, denoted as T0 = T0 + Ta (where “=" means assignment).

[0124] After determining the maximum amount of traffic that a network node needs to allocate in the future, the traffic can be allocated based on the maximum amount, thereby reducing the scheduling frequency and completing the scheduling in one go.

[0125] Existing scheduling methods rely on current data, which may result in insufficient scheduling capacity in a single operation, necessitating future scheduling for the same location and leading to high scheduling frequency. This application's embodiment utilizes predicted future data for scheduling, directly scheduling according to the maximum required value, thus reducing scheduling frequency.

[0126] In some embodiments, the number of network nodes is at least two.

[0127] Traffic scheduling based on inbound and outbound traffic includes: classifying network nodes based on inbound and outbound traffic for each network node to obtain at least one inbound traffic node and at least one outbound traffic node; determining the inbound traffic node corresponding to each outbound traffic node based on the outbound traffic of each outbound traffic node and the traffic utilization rate of each inbound traffic node; and scheduling the outbound traffic of each outbound traffic node to the corresponding inbound traffic node.

[0128] Specifically, network nodes are classified based on the incoming and outgoing traffic of each network node to obtain at least one traffic-incoming node and at least one traffic-outgoing node.

[0129] Furthermore, based on the required outgoing traffic of each outgoing traffic node and the traffic utilization rate of each incoming traffic node, the corresponding incoming traffic node for each outgoing traffic node is determined; the required outgoing traffic of each outgoing traffic node is then scheduled to the corresponding incoming traffic node.

[0130] Specifically, all traffic outgoing nodes are sorted from largest to smallest according to the amount of traffic to be outgoing, and all traffic incoming nodes are sorted from smallest to largest according to traffic utilization.

[0131] Furthermore, the traffic that needs to be transferred out of the node with the largest traffic volume is scheduled to the traffic that needs to be transferred in with the lowest traffic utilization.

[0132] After completing the traffic scheduling of a traffic outgoing node, update the traffic utilization of each traffic incoming node and the traffic to be dispatched by each traffic outgoing node, and reorder them to proceed with the traffic scheduling of the next traffic outgoing node, until all traffic scheduling is completed.

[0133] In some embodiments, after traffic scheduling is performed based on inbound traffic and outbound traffic, the method further includes: calculating scheduling compensation for CDN predicted traffic, scheduling compensation for switch predicted traffic, and scheduling compensation for metropolitan area network predicted traffic based on the least squares method.

[0134] Specifically, the scheduling compensation for CDN predicted traffic is used to compensate for the CDN predicted traffic output by the traffic prediction model in the next prediction period; the scheduling compensation for switch predicted traffic is used to compensate for the switch predicted traffic output by the traffic prediction model in the next prediction period; and the scheduling compensation for metropolitan area network predicted traffic is used to compensate for the metropolitan area network predicted traffic output by the traffic prediction model in the next prediction period.

[0135] Specifically, such as Figure 3As shown, since the scheduling process itself has a certain delay, in order to reflect the effect of the current scheduling process in the future, after traffic scheduling based on the traffic that can be scheduled in and the traffic that needs to be scheduled out, the actual scheduling situation can be modeled according to the least squares method to obtain the scheduling compensation for CDN predicted traffic, switch predicted traffic and metropolitan area network predicted traffic in the current scheduling process. The calculated scheduling compensation is used to compensate for the CDN predicted traffic, switch predicted traffic and metropolitan area network predicted traffic output by the traffic prediction model in the next prediction period, so as to improve the accuracy of the prediction results.

[0136] Furthermore, such as Figure 2 As shown, after the intelligent scheduling module on the server side completes traffic scheduling, it can send the scheduling results to the client for display.

[0137] This application also provides a traffic scheduling device; please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic diagram of the traffic scheduling device provided in an embodiment of this application. In this embodiment, the traffic scheduling device includes an acquisition module 510, a prediction module 520, a determination module 530, and a scheduling module 540.

[0138] The acquisition module 510 is used to acquire historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data of network nodes.

[0139] The prediction module 520 is used to input CDN historical traffic data, switch historical traffic data and metropolitan area network historical traffic data into the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic and metropolitan area network predicted traffic at future moments output by the traffic prediction model.

[0140] The determination module 530 is used to determine the inbound and outbound traffic of network nodes based on CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic.

[0141] The scheduling module 540 is used for traffic scheduling based on the traffic that can be scheduled in and the traffic that needs to be scheduled out.

[0142] The traffic prediction model is trained based on sample CDN historical traffic data and corresponding sample CDN predicted traffic, sample switch historical traffic data and corresponding sample switch predicted traffic, and sample metropolitan area network historical traffic data and corresponding sample metropolitan area network predicted traffic.

[0143] In some embodiments, the traffic prediction model includes a decomposition network and an additive network. The decomposition network decomposes CDN historical traffic data, switch historical traffic data, and metropolitan area network (MAN) historical traffic data to generate trend components, periodic components, holiday components, and error components. The additive network fits and predicts the trend components, periodic components, holiday components, and error components to generate predicted CDN traffic, switch traffic, and MAN traffic for future time periods. The trend components characterize the non-periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components characterize the periodic variation trend of CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The periodic components are represented by Fourier series. The holiday components characterize the impact of non-periodic holidays on CDN historical traffic data, switch historical traffic data, and MAN historical traffic data. The error components characterize unpredictable fluctuations in the model.

[0144] In some embodiments, the prediction module 520 is used to input CDN historical traffic data, switch historical traffic data, and metropolitan area network historical traffic data into the decomposition network of the traffic prediction model to obtain the trend component, periodic component, holiday component, and error component output by the decomposition network; and to input the trend component, periodic component, holiday component, and error component into the addition network of the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times output by the addition network.

[0145] In some embodiments, the prediction module 520 is used to determine the data mutation compensation for CDN predicted traffic, the data mutation compensation for switch predicted traffic, and the data mutation compensation for metropolitan area network predicted traffic; and to perform superposition calculations on the data mutation compensation and CDN predicted traffic, the data mutation compensation and switch predicted traffic, and the data mutation compensation and metropolitan area network predicted traffic, respectively, to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times after compensation processing.

[0146] In some embodiments, the determining module 530 is used to determine whether a network node meets the scheduling conditions based on CDN predicted traffic and switch predicted traffic; if the network node meets the scheduling conditions, then it determines CDN inbound traffic and CDN outbound traffic based on CDN predicted traffic and CDN traffic security line, determines switch inbound traffic and switch outbound traffic based on switch predicted traffic and switch traffic security line, and determines metropolitan area network inbound traffic and metropolitan area network outbound traffic based on metropolitan area network predicted traffic and metropolitan area network maximum traffic value; it determines the network node's available inbound traffic based on CDN inbound traffic, switch inbound traffic, and metropolitan area network inbound traffic; and it determines the network node's required outbound traffic based on CDN outbound traffic, switch outbound traffic, and metropolitan area network outbound traffic.

[0147] In some embodiments, the number of network nodes is at least two.

[0148] The scheduling module 540 is used to classify network nodes based on the inbound traffic and outbound traffic of each network node, and obtain at least one traffic inbound node and at least one traffic outbound node; based on the outbound traffic of each traffic outbound node and the traffic utilization rate of each traffic inbound node, determine the traffic inbound node corresponding to each traffic outbound node; and schedule the outbound traffic of each traffic outbound node to the corresponding traffic inbound node.

[0149] In some embodiments, the prediction module 520 is used to calculate the scheduling compensation for CDN predicted traffic, the scheduling compensation for switch predicted traffic, and the scheduling compensation for metropolitan area network predicted traffic based on the least squares method; wherein, the scheduling compensation for CDN predicted traffic is used to compensate the CDN predicted traffic output by the traffic prediction model for the next prediction period; the scheduling compensation for switch predicted traffic is used to compensate the switch predicted traffic output by the traffic prediction model for the next prediction period; and the scheduling compensation for metropolitan area network predicted traffic is used to compensate the metropolitan area network predicted traffic output by the traffic prediction model for the next prediction period.

[0150] This application also provides an electronic device. Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 6 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions from the memory 630 to execute a flow scheduling method.

[0151] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0152] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the traffic scheduling methods provided by the methods described above.

[0153] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0154] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method of flow scheduling, the method comprising: include: Obtain historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data of network nodes; The CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data are input into the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times output by the traffic prediction model. Based on the CDN predicted traffic, the switch predicted traffic, and the metropolitan area network predicted traffic, the inbound traffic and outbound traffic of the network node are determined. Traffic scheduling is performed based on the incoming traffic and the outgoing traffic. The traffic prediction model is trained based on sample CDN historical traffic data and the sample CDN predicted traffic corresponding to the sample CDN historical traffic data, sample switch historical traffic data and the sample switch predicted traffic corresponding to the sample switch historical traffic data, and sample metropolitan area network historical traffic data and the sample metropolitan area network predicted traffic corresponding to the sample metropolitan area network historical traffic data. The process of determining the inbound and outbound traffic of the network node based on the CDN predicted traffic, the switch predicted traffic, and the metropolitan area network predicted traffic includes: Based on the CDN predicted traffic and the switch predicted traffic, it is determined whether the network node meets the scheduling conditions; If the network node meets the scheduling conditions, then the CDN incoming traffic and CDN outgoing traffic are determined based on the CDN predicted traffic and CDN traffic safety line, the switch incoming traffic and switch outgoing traffic are determined based on the switch predicted traffic and switch traffic safety line, and the metropolitan area network incoming traffic and metropolitan area network outgoing traffic are determined based on the metropolitan area network predicted traffic and metropolitan area network maximum traffic value. Based on CDN inbound traffic, switch inbound traffic, and metropolitan area network inbound traffic, the available inbound traffic of the network node is determined; Based on CDN outbound traffic, switch outbound traffic, and metropolitan area network outbound traffic, the required outbound traffic of the network node is determined.

2. The traffic scheduling method of claim 1, wherein, The traffic prediction model includes a decomposition network and an additive network; The decomposition network is used to decompose the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data to generate trend components, periodic components, holiday components, and error components. The additive network is used to fit and predict the trend component, the periodic component, the holiday component, and the error component to generate CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic for future times. The trend component is used to characterize the non-periodic change trend of the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data. The periodic component is used to characterize the periodic change trend of the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data; the periodic component is represented by a Fourier series. The holiday component is used to characterize the impact of non-periodic holidays on the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data; The error components are used to characterize unpredictable fluctuations in the model.

3. The traffic scheduling method according to claim 2, characterized in that, The step of inputting the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data into the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times output by the traffic prediction model includes: The CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data are input into the decomposition network of the traffic prediction model to obtain the trend component, periodic component, holiday component, and error component output by the decomposition network. The trend component, the periodic component, the holiday component, and the error component are input into the additive network of the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future moments output by the additive network.

4. The traffic scheduling method according to claim 1, characterized in that, Before determining the inbound and outbound traffic of the network node based on the CDN predicted traffic, the switch predicted traffic, and the metropolitan area network predicted traffic, the method further includes: Determine the data mutation compensation for the CDN predicted traffic, the data mutation compensation for the switch predicted traffic, and the data mutation compensation for the metropolitan area network predicted traffic; The data mutation compensation and CDN predicted traffic, the data mutation compensation and switch predicted traffic, and the data mutation compensation and metropolitan area network predicted traffic are respectively superimposed and calculated to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times after compensation processing.

5. The traffic scheduling method according to claim 1, characterized in that, The number of network nodes is at least two; The traffic scheduling based on the inbound traffic and the outbound traffic includes: Based on the incoming and outgoing traffic of each network node, network nodes are classified to obtain at least one traffic-incoming node and at least one traffic-outgoing node. Based on the required outgoing traffic of each outgoing traffic node and the traffic utilization rate of each incoming traffic node, determine the incoming traffic node corresponding to each outgoing traffic node. The traffic to be dispatched from each traffic dispatching node is dispatched to the corresponding traffic dispatching node.

6. The traffic scheduling method according to claim 1, characterized in that, After performing traffic scheduling based on the incoming traffic and the outgoing traffic, the method further includes: The scheduling compensation for the CDN predicted traffic, the scheduling compensation for the switch predicted traffic, and the scheduling compensation for the metropolitan area network predicted traffic are calculated based on the least squares method. The CDN predicted traffic scheduling compensation is used to compensate the CDN predicted traffic output by the traffic prediction model in the next prediction period. The scheduling compensation for the switch predicted traffic is used to compensate the switch predicted traffic output by the traffic prediction model in the next prediction period. The scheduling compensation for the metropolitan area network (MAN) predicted traffic is used to compensate for the MAN predicted traffic output by the traffic prediction model in the next prediction period.

7. A flow scheduling device, characterized in that, include: The acquisition module is used to acquire historical CDN traffic data, historical switch traffic data, and historical metropolitan area network traffic data of network nodes. The prediction module is used to input the CDN historical traffic data, the switch historical traffic data, and the metropolitan area network historical traffic data into the traffic prediction model to obtain the CDN predicted traffic, switch predicted traffic, and metropolitan area network predicted traffic at future times output by the traffic prediction model. The determination module is used to determine the inbound traffic and outbound traffic of the network node based on the CDN predicted traffic, the switch predicted traffic, and the metropolitan area network predicted traffic. The scheduling module is used to perform traffic scheduling based on the inbound traffic and the outbound traffic; The traffic prediction model is trained based on sample CDN historical traffic data and the sample CDN predicted traffic corresponding to the sample CDN historical traffic data, sample switch historical traffic data and the sample switch predicted traffic corresponding to the sample switch historical traffic data, and sample metropolitan area network historical traffic data and the sample metropolitan area network predicted traffic corresponding to the sample metropolitan area network historical traffic data. The determining module is used to determine whether the network node meets the scheduling conditions based on the CDN predicted traffic and the switch predicted traffic; if the network node meets the scheduling conditions, then it determines the CDN inbound traffic and CDN outbound traffic based on the CDN predicted traffic and CDN traffic safety line, determines the switch inbound traffic and switch outbound traffic based on the switch predicted traffic and switch traffic safety line, and determines the metropolitan area network inbound traffic and metropolitan area network outbound traffic based on the metropolitan area network predicted traffic and the maximum metropolitan area network traffic; it determines the network node's available inbound traffic based on the CDN inbound traffic, switch inbound traffic, and metropolitan area network inbound traffic; and it determines the network node's required outbound traffic based on the CDN outbound traffic, switch outbound traffic, and metropolitan area network outbound traffic.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the traffic scheduling method as described in any one of claims 1 to 6.

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

Images