SD-WAN application level QoE real-time monitoring and scheduling method and system

Abstract

The invention discloses an SD-WAN application level QoE (Quality of Experience) real-time monitoring and scheduling method and system, relates to the technical field of networks, and solves the problems of insufficient collaboration of application identification and path scheduling, inaccurate QoE monitoring, lagging scheduling decision, large resource consumption and much manual intervention in the prior art. Comprising the following steps: step 1, constructing an SD-WAN application-level multi-dimensional monitoring system, and collecting an application-level QoE index and a link-level QoS index by adopting a mode of combining active detection and passive monitoring; according to the method, a multi-dimensional monitoring mode combining active detection and passive monitoring is adopted, full-dimensional indexes of application QoE and link QoS are covered, the accuracy of monitoring data is improved by combining data preprocessing, and the problems of incomplete monitoring coverage and inaccurate data in the prior art are solved; and a correlation model of the application QoE and the link QoS is established, and the influence weight of each QoS index on the QoE is quantified, so that the QoE evaluation is more suitable for the actual operation state of the application.

Description

Technical Field

[0001] This invention relates to the field of network technology, specifically to a method and system for real-time monitoring and scheduling of SD-WAN application-level QoE. Background Technology

[0002] SD-WAN, as a next-generation wide area network technology, has been widely applied in enterprise cross-regional networking and cloud-edge collaboration scenarios due to its flexible networking methods and intelligent path selection capabilities. In practical applications of SD-WAN, application-level QoE is a core indicator for measuring network service quality and directly impacts the user's service experience.

[0003] In existing technologies, SD-WAN application identification and path scheduling suffer from significant coordination deficiencies: most solutions can only achieve simple application port / protocol classification, employing fixed path allocation and bandwidth configuration strategies. They cannot dynamically adjust routing strategies based on real-time changes in application QoE, resulting in significant lag in scheduling decisions when link latency increases or packet loss rates rise. Simultaneously, existing monitoring solutions have several limitations: first, they employ a single monitoring method, often using passive monitoring or single active probes, leading to incomplete QoE monitoring coverage, low data accuracy, and an inability to accurately reflect the actual operational experience of applications; second, monitoring data processing efficiency is low, lacking correlation analysis between application QoE and link QoS (Quality of Service), making real-time data analysis difficult in large-scale SD-WAN application scenarios; third, the monitoring frequency is fixed, using the same probe frequency regardless of network status and application QoE changes, resulting in significant unnecessary network resource consumption. Furthermore, existing scheduling systems rely on manual intervention, lacking automated scheduling mechanisms and visualized management platforms, failing to achieve full-process monitoring and rapid response in the scheduling process.

[0004] To address the above problems, this invention provides a method and system for real-time monitoring and scheduling of SD-WAN application-level QoE. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for real-time monitoring and scheduling of application-level QoE in SD-WAN, solving the problems of insufficient coordination between application identification and path scheduling, inaccurate QoE monitoring, delayed scheduling decisions, high resource consumption, and excessive manual intervention in existing technologies.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring and scheduling method and system for SD-WAN application-level QoE, comprising the following steps:

[0007] Step 1: Construct a multi-dimensional monitoring system for SD-WAN applications. This system combines active detection with passive monitoring to collect application-level QoE metrics and link-level QoS metrics. The collected metrics data are then preprocessed. Active detection uses ICMP, TCPSYN, and application-layer simulated requests to detect link latency, packet loss rate, and jitter.

[0008] Passive monitoring collects metrics such as application throughput, response time, and session success rate through SD-WAN edge nodes;

[0009] QoS metrics include link latency (D), packet loss rate (L), jitter (J), and available bandwidth (B); QoE metrics include application response time (T), throughput (TP), session success rate (S), and number of stutters (C).

[0010] Preprocessing methods include data cleaning, noise reduction, and normalization.

[0011] Step 2: Establish a correlation model between application QoE and link QoS. Determine the influence weight of link QoS indicators on application QoE through correlation coefficient calculation. Calculate the real-time application QoE value based on the preprocessed indicator data to achieve accurate evaluation of application QoE. The formula for calculating the real-time application QoE value is: QoE = k1×T + k2×TP + k3×S + k4×C, where k1, k2, k3, and k4 are the weight coefficients of the QoE indicators, and k1 + k2 + k3 + k4 = 1. The weight coefficients are dynamically adjusted according to the application type.

[0012] The correlation coefficient between QoE and link QoS is calculated as follows: R = (QoE change rate / QoS indicator change rate) × 100%. The influence weights of each QoS indicator, ωD, ωL, ωJ, and ωB, are determined based on the magnitude of the correlation coefficient, and ωD + ωL + ωJ + ωB = 1.

[0013] Step 3: Based on the AI ​​prediction model, predict the application QoE change trend, and combine the real-time QoE value to make anomaly judgment, classifying QoE into three levels: normal, warning, and fault. The AI ​​prediction model adopts the LSTM time series prediction model, using historical QoE index data as the training set, and outputs the application QoE prediction value within the future preset time period.

[0014] The anomaly judgment rules are as follows: when the real-time QoE value is within [QoE standard value × 90%, QoE standard value × 110%], it is at the normal level; when the real-time QoE value is within [QoE standard value × 70%, QoE standard value × 90%) or (QoE standard value × 110%, QoE standard value × 130%], it is at the warning level; and when the real-time QoE value is <QoE standard value × 70% or >QoE standard value × 130%, it is at the fault level.

[0015] Step 4: Develop dynamic scheduling strategies for different QoE levels, calculate the scheduling adaptation value of each candidate link using a multi-objective optimization algorithm, and automate link selection, bandwidth allocation, and link switching based on the adaptation value. The formula for calculating the scheduling adaptation value of a candidate link using the multi-objective optimization algorithm is: F=ωD×(1-D / Dmax)+ωL×(1-L / Lmax)+ωJ×(1-J / Jmax)+ωB×(B / Bmax), where Dmax, Lmax, and Jmax are the threshold values ​​of the link's QoS indicators, and Bmax is the maximum available bandwidth of the link. The scheduling strategy is as follows: for normal levels, maintain the current link and bandwidth configuration; for warning levels, dynamically allocate bandwidth; and for fault levels, select the candidate link with the largest scheduling adaptation value for seamless switching.

[0016] Step 5: Dynamically adjust the monitoring frequency based on the application QoE level and network load status. Simultaneously, display the monitoring data and scheduling results through a visual management platform to achieve full-process monitoring and manual intervention assistance during the scheduling process. The dynamic adjustment rules for the monitoring frequency are as follows: When the application QoE is at a normal level and the network load is <50%, the monitoring frequency is F0; when the application QoE is at a warning level or the network load is between 50% and 80%, the monitoring frequency is 2×F0; when the application QoE is at a fault level or the network load is >80%, the monitoring frequency is 5×F0, with F0 being the base monitoring frequency.

[0017] An SD-WAN application-level QoE real-time monitoring and scheduling system includes: a multi-dimensional data acquisition module, a QoE evaluation and correlation module, an AI prediction and anomaly judgment module, a dynamic scheduling decision module, a monitoring frequency adaptive module, and a visualization management module;

[0018] The multi-dimensional data acquisition module is used to collect application QoE and link QoS indicators of SD-WAN through active detection and passive monitoring, and to complete data preprocessing.

[0019] The QoE assessment and correlation module is used to establish a correlation model between application QoE and link QoS, calculate the correlation coefficient and influence weight, and output the real-time application QoE value.

[0020] The AI ​​prediction and anomaly detection module is used to predict the QoE change trend based on the LSTM model and classify the level by combining the real-time QoE value.

[0021] The dynamic scheduling decision module is used to formulate scheduling strategies based on QoE level, calculate link adaptation value through multi-objective optimization algorithm, and realize the automation of link selection, bandwidth allocation and switching.

[0022] The monitoring frequency adaptive module is used to dynamically adjust the monitoring frequency of data acquisition according to the application QoE level and network load.

[0023] The visualization management module is used to display monitoring indicators, QoE levels, and scheduling results, and provides a human-computer interaction interface to enable manual intervention.

[0024] Preferably, the multi-dimensional data acquisition module includes an active detection unit and a passive monitoring unit. The active detection unit is deployed at the SD-WAN control node, and the passive monitoring unit is deployed at each SD-WAN edge node. The edge nodes and the control node interact with each other through a dedicated channel.

[0025] Preferably, the dynamic scheduling decision module has a built-in application type library, which can dynamically adjust the QoE indicator weight coefficients k1, k2, k3, and k4 according to different application types such as voice, video, data transmission, and office systems.

[0026] Preferably, the visualization management module supports real-time data dashboards, historical data tracing, scheduling log querying, and abnormal alarm push notifications. Alarm push notification methods include SMS, email, and platform pop-ups.

[0027] This invention provides a method and system for real-time monitoring and scheduling of SD-WAN application-level QoE. Compared with existing technologies, it has the following advantages:

[0028] This invention employs a multi-dimensional monitoring approach that combines active detection with passive monitoring, covering all dimensions of application QoE and link QoS indicators. Combined with data preprocessing, it improves the accuracy of monitoring data and solves the problems of incomplete monitoring coverage and inaccurate data in existing technologies.

[0029] Establish a correlation model between application QoE and link QoS, quantify the impact weight of each QoS indicator on QoE, so that QoE evaluation is more in line with the actual operation status of the application, and provide accurate data support for scheduling decisions.

[0030] By combining AI prediction models and real-time monitoring data to identify QoE anomalies, the trend of QoE changes can be predicted in advance, enabling early detection and handling of anomalies and solving the problem of delayed scheduling decisions.

[0031] Differentiated dynamic scheduling strategies are formulated based on different QoE levels. Link adaptation values ​​are calculated by combining multi-objective optimization algorithms to automate link selection, bandwidth allocation and switching, thereby improving the rationality and real-time performance of scheduling decisions. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0033] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] First Embodiment

[0036] SD-WAN Application-Level QoE Real-Time Monitoring and Scheduling Method

[0037] The SD-WAN application-level QoE real-time monitoring and scheduling method disclosed in this embodiment specifically includes the following steps: Step 1: Constructing an SD-WAN application-level multi-dimensional monitoring system, collecting and preprocessing indicator data, and constructing a full-dimensional monitoring system by combining active detection and passive monitoring.

[0038] Active Probe: Initiated by the SD-WAN control node, it uses ICMP to probe link latency (D) and packet loss rate (L), TCPSYN to probe link jitter (J), and application layer simulation requests to probe available link bandwidth (B). The probe range covers all SD-WAN backbone links and access links.

[0039] Passive monitoring: Each SD-WAN edge node collects application traffic passing through it to obtain application-level QoE metrics, including application response time T, throughput TP, session success rate S, and number of stutters C (for real-time video and voice applications).

[0040] The collected QoS and QoE index data are preprocessed as follows: First, outliers and missing values ​​are removed by data cleaning; then, wavelet transform is used to denoise the data and eliminate errors caused by network interference; finally, normalization is performed to map all index data to the [0,1] interval for easy subsequent calculation.

[0041] Step 2: Establish a QoE and QoS correlation model and calculate the application's real-time QoE value.

[0042] Calculate the real-time QoE value of the application: Dynamically adjust the weight coefficients k1, k2, k3, and k4 of the QoE indicator (the weights sum to 1) according to the application type (voice, video, office data, etc.). For example, video applications have a higher weight coefficient k4 for the number of stutters C, while office data applications have a higher weight coefficient k2 for throughput TP. The real-time QoE value is calculated using the formula: QoE = k1 × T + k2 × TP + k3 × S + k4 × C. Wherein, the application response time T and the number of stutters C are negative indicators and need to be inverted first to ensure that the larger the indicator value, the higher its contribution to QoE.

[0043] Establish a QoE and QoS correlation model: Calculate the correlation coefficient between each link's QoS indicator and the application's QoE. The correlation coefficient formula is: R = (QoE change rate / QoS indicator change rate) × 100%. Based on the magnitude of the correlation coefficient R, determine the influence weights ωD, ωL, ωJ, and ωB (the sum of the weights is 1) of each QoS indicator on the application's QoE. The larger the correlation coefficient, the higher the influence weight of the corresponding QoS indicator. For example, in video applications, the link latency D has the largest correlation coefficient, and its influence weight ωD is the highest.

[0044] Step 3: AI Prediction and QoE Anomaly Level Judgment

[0045] AI Prediction: The LSTM time series prediction model is used. The model is trained with historical QoE data from the past hour as the training set and outputs the application QoE prediction value for the next 5 minutes to predict the QoE trend in advance.

[0046] Anomaly detection: Based on the application's business requirements, a standard QoE value is set, and combined with real-time QoE values ​​and predicted values, a grading system is established, defining three levels: normal, warning, and fault.

[0047] Normal level: The real-time QoE value is within [QoE standard value × 90%, QoE standard value × 110%], and the predicted value does not show a significant downward trend;

[0048] Warning level: The real-time QoE value is at [70% of the standard QoE value, 90% of the standard QoE value] or (110% of the standard QoE value, 130% of the standard QoE value], or the predicted value shows a downward trend;

[0049] Fault level: Real-time QoE value < 70% of standard QoE value or > 130% of standard QoE value, or predicted value will be lower than 70% of standard QoE value.

[0050] Step 4: Execution of Dynamic Scheduling Strategies Based on Multi-Objective Optimization Algorithms. Differentiated dynamic scheduling strategies are formulated for different QoE levels, and scheduling adaptation values ​​for candidate links are calculated using multi-objective optimization algorithms to achieve automated scheduling.

[0051] Calculate the scheduling adaptation value: Combining the QoS indicator influence weights obtained in step two, with the goal of minimizing link latency, packet loss rate, jitter, and maximizing available bandwidth, calculate the scheduling adaptation value for each candidate link: F=ωD×(1-D / Dmax)+ωL×(1-L / Lmax)+ωJ×(1-J / Jmax)+ωB×(B / Bmax) where Dmax, Lmax, and Jmax are pre-set QoS indicator thresholds for the link. If the threshold is exceeded, the link will not participate in scheduling; Bmax is the maximum available bandwidth of the link. The adaptation value F ranges from [0,1], and the larger the value, the more suitable the link is for carrying the current application.

[0052] Differentiated scheduling strategy:

[0053] Normal level: Maintain the current link selection and bandwidth configuration, and only perform real-time monitoring;

[0054] Warning level: No link switching will be performed; only the bandwidth of the current link will be dynamically allocated to prioritize the bandwidth requirements of this application, while closely monitoring QoE changes.

[0055] Fault Level: Immediately initiate link switching, select the candidate link with the largest scheduling adaptation value F to achieve seamless switching, and at the same time mark the original faulty link and notify the operation and maintenance personnel to investigate.

[0056] Step 5: Dynamic Adjustment and Visual Management of Monitoring Frequency

[0057] Dynamic adjustment of monitoring frequency: Set a basic monitoring frequency F0 (e.g., 1 time / minute), and dynamically adjust it according to the application QoE level and network load status, with the following rules:

[0058] Normal level + network load < 50%: monitoring frequency = F0;

[0059] Warning level or network load 50%-80%: Monitoring frequency = 2 × F0;

[0060] Fault level or network load > 80%: Monitoring frequency = 5 × F0. This reduces detection overhead when network load is low and QoE is normal, and increases monitoring density under abnormal conditions.

[0061] Visualization and Automation Management: The visualization management platform displays real-time QoE metrics, QoS metrics, QoE levels, link scheduling status, and other information, supporting historical data tracing and scheduling log querying. It also sets up anomaly alarm push notifications, sending warnings and fault information to operations and maintenance personnel via SMS, email, and platform pop-ups. The platform provides a human-computer interaction interface, allowing operations and maintenance personnel to provide manual intervention based on actual conditions, achieving a combination of automated scheduling and manual intervention.

[0062] Second Embodiment

[0063] SD-WAN Application-Level QoE Real-Time Monitoring and Scheduling System

[0064] The SD-WAN application-level QoE real-time monitoring and scheduling system disclosed in this embodiment is used to implement the method in Embodiment 1. This system is deployed on an SD-WAN control platform and includes a multi-dimensional data acquisition module, a QoE assessment and correlation module, an AI prediction and anomaly detection module, a dynamic scheduling decision module, a monitoring frequency adaptive module, and a visualization management module. These modules work collaboratively to achieve real-time monitoring and automated scheduling of application-level QoE. The specific functions of each module are as follows:

[0065] The multi-dimensional data acquisition module includes an active detection unit and a passive monitoring unit. The active detection unit is deployed at the SD-WAN control node to perform active detection tasks such as ICMP and TCPSYN. The passive monitoring unit is deployed at each SD-WAN edge node to collect the QoE index of application traffic. After completing the data acquisition, this module performs preprocessing such as cleaning, noise reduction, and normalization on the data, and transmits the preprocessed data to the subsequent modules through a dedicated channel.

[0066] QoE Assessment and Correlation Module: With a built-in application type library, it can dynamically adjust the weight coefficients of QoE indicators according to different application types such as voice, video, and office data. This module receives pre-processed indicator data, calculates the real-time QoE value of the application, establishes a correlation model between QoE and QoS, quantifies the influence weight of each QoS indicator, and transmits the calculation results to the AI ​​prediction and anomaly judgment module.

[0067] AI Prediction and Anomaly Detection Module: It has a built-in trained LSTM prediction model to predict the application's QoE change trend; at the same time, it judges the anomaly level based on the preset QoE standard value, combined with the real-time QoE value and the predicted value, and transmits the QoE level result to the dynamic scheduling decision module and the monitoring frequency adaptive module.

[0068] Dynamic scheduling decision module: Receives QoE level results and executes differentiated scheduling strategies according to different levels; it has a built-in multi-objective optimization algorithm to calculate the scheduling adaptation value of each candidate link and automatically complete link selection, bandwidth allocation and seamless link switching; at the same time, this module can interact with SD-WAN edge nodes in real time and issue scheduling instructions.

[0069] Monitoring frequency adaptive module: Receives QoE level results and collects real-time load status of the SD-WAN network. It dynamically adjusts the monitoring frequency of the multi-dimensional data acquisition module according to preset rules to reduce network resource consumption.

[0070] Visual management module: As the core of human-computer interaction, it supports real-time data dashboards, historical data tracing, and scheduling log querying, intuitively displaying monitoring and scheduling-related information; it also provides anomaly alarm push function, supporting three push methods: SMS, email, and platform pop-up, and provides a manual intervention interface, where operation and maintenance personnel can issue manual scheduling instructions to achieve the collaboration between automation and manual assistance.

[0071] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0072] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A real-time monitoring and scheduling method for SD-WAN application-level QoE, characterized in that, Includes the following steps: Step 1: Construct a multi-dimensional monitoring system for SD-WAN applications, using a combination of active detection and passive monitoring to collect application-level QoE metrics and link-level QoS metrics, and preprocess the collected metric data. Step 2: Establish a correlation model between application QoE and link QoS, determine the influence weight of link QoS indicators on application QoE by calculating the correlation coefficient, and calculate the real-time application QoE value based on the preprocessed indicator data to achieve accurate evaluation of application QoE. Step 3: Based on the AI ​​prediction model, predict the application QoE change trend, combine the real-time QoE value to make anomaly judgment, and classify QoE into three levels: normal, warning, and fault. Step 4: Develop dynamic scheduling strategies for different QoE levels, calculate the scheduling adaptation value of each candidate link using a multi-objective optimization algorithm, and complete the automated scheduling of link selection, bandwidth allocation, and link switching based on the adaptation value. Step 5: Dynamically adjust the monitoring frequency based on the application QoE level and network load status, and display the monitoring data and scheduling results through a visualization management platform to achieve full-process monitoring and manual intervention assistance in the scheduling process.

2. The SD-WAN application-level QoE real-time monitoring and scheduling method according to claim 1, characterized in that, In step one, the active detection uses ICMP, TCPSYN and application layer simulated requests to detect link latency, packet loss rate and jitter. Passive monitoring collects metrics such as application throughput, response time, and session success rate through SD-WAN edge nodes; QoS metrics include link latency (D), packet loss rate (L), jitter (J), and available bandwidth (B); QoE metrics include application response time (T), throughput (TP), session success rate (S), and number of stutters (C). Preprocessing methods include data cleaning, noise reduction, and normalization.

3. The SD-WAN application-level QoE real-time monitoring and scheduling method according to claim 1, characterized in that, In step two, the formula for calculating the real-time QoE value is: QoE=k1×T+k2×TP+k3×S+k4×C, where k1, k2, k3, and k4 are the weight coefficients of the QoE index, and k1+k2+k3+k4=1. The weight coefficients are dynamically adjusted according to the application type. The correlation coefficient between QoE and link QoS is calculated as follows: R = (QoE change rate / QoS indicator change rate) × 100%. The influence weights of each QoS indicator, ωD, ωL, ωJ, and ωB, are determined based on the magnitude of the correlation coefficient, and ωD + ωL + ωJ + ωB = 1.

4. The SD-WAN application-level QoE real-time monitoring and scheduling method according to claim 1, characterized in that, In step three, the AI ​​prediction model uses the LSTM time series prediction model, with historical QoE index data as the training set, to output the application QoE prediction value within a preset time period in the future. The anomaly judgment rules are as follows: when the real-time QoE value is within [QoE standard value × 90%, QoE standard value × 110%], it is at the normal level; when the real-time QoE value is within [QoE standard value × 70%, QoE standard value × 90%) or (QoE standard value × 110%, QoE standard value × 130%], it is at the warning level; and when the real-time QoE value is <QoE standard value × 70% or >QoE standard value × 130%, it is at the fault level.

5. The SD-WAN application-level QoE real-time monitoring and scheduling method according to claim 1, characterized in that, In step four, the multi-objective optimization algorithm calculates the candidate link scheduling adaptation value using the following formula: F = ωD × (1 - D / Dmax) + ωL × (1 - L / Lmax) + ωJ × (1 - J / Jmax) + ωB × (B / Bmax), where Dmax, Lmax, and Jmax are the threshold values ​​of the link QoS indicators, and Bmax is the maximum available bandwidth of the link. The scheduling strategy is as follows: for normal levels, the current link and bandwidth configuration are maintained; for warning levels, bandwidth is dynamically allocated; and for fault levels, the candidate link with the largest scheduling adaptation value is selected for seamless switching.

6. The SD-WAN application-level QoE real-time monitoring and scheduling method according to claim 1, characterized in that, In step five, the dynamic adjustment rule for the monitoring frequency is as follows: when the application QoE is at the normal level and the network load is <50%, the monitoring frequency is F0; when the application QoE is at the warning level or the network load is between 50% and 80%, the monitoring frequency is 2×F0; when the application QoE is at the fault level or the network load is >80%, the monitoring frequency is 5×F0, with F0 being the basic monitoring frequency.

7. An SD-WAN application-level QoE real-time monitoring and scheduling system, used to implement the SD-WAN application-level QoE real-time monitoring and scheduling method according to any one of claims 1-6, characterized in that, include: Multi-dimensional data acquisition module, QoE assessment and correlation module, AI prediction and anomaly detection module, dynamic scheduling decision module, monitoring frequency adaptive module, and visualization management module; The multi-dimensional data acquisition module is used to collect application QoE and link QoS indicators of SD-WAN through active detection and passive monitoring, and to complete data preprocessing. The QoE assessment and correlation module is used to establish a correlation model between application QoE and link QoS, calculate the correlation coefficient and influence weight, and output the real-time application QoE value. The AI ​​prediction and anomaly detection module is used to predict the QoE change trend based on the LSTM model and classify the level by combining the real-time QoE value. The dynamic scheduling decision module is used to formulate scheduling strategies based on QoE level, calculate link adaptation value through multi-objective optimization algorithm, and realize the automation of link selection, bandwidth allocation and switching. The monitoring frequency adaptive module is used to dynamically adjust the monitoring frequency of data acquisition according to the application QoE level and network load. The visualization management module is used to display monitoring indicators, QoE levels, and scheduling results, and provides a human-computer interaction interface to enable manual intervention.

8. The SD-WAN application-level QoE real-time monitoring and scheduling system according to claim 7, characterized in that, The multi-dimensional data acquisition module includes an active detection unit and a passive monitoring unit. The active detection unit is deployed at the SD-WAN control node, and the passive monitoring unit is deployed at each SD-WAN edge node. The edge nodes and the control node interact with each other through a dedicated channel.

9. The SD-WAN application-level QoE real-time monitoring and scheduling system according to claim 7, characterized in that, The dynamic scheduling decision module has a built-in application type library, which can dynamically adjust the QoE indicator weight coefficients k1, k2, k3, and k4 according to different application types such as voice, video, data transmission, and office systems.

10. The SD-WAN application-level QoE real-time monitoring and scheduling system according to claim 7, characterized in that, The visualization management module supports real-time data dashboards, historical data tracing, scheduling log querying, and anomaly alarm push notifications. Alarm push notification methods include SMS, email, and platform pop-ups.

Images