A method and system for monitoring the testing process of computer network application programs

Abstract

The present invention relates to a method and system for monitoring the test process of a computer network application program, including creating an isolated test environment based on virtualization technology, integrating multiple data sources during the test process, using an improved deep learning algorithm to analyze real-time data, identifying potential abnormal behaviors and performance bottlenecks, and displaying the test results to testers through visualization tools. According to the test results in intelligent feedback, the computing resources and network resources in the test environment are automatically adjusted to ensure resource utilization and system stability during the test process. The present invention uses an improved deep learning algorithm to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and use an adaptive model update mechanism to continuously optimize the detection model. Through the combination of machine learning and deep learning technology, the system can dynamically adapt to complex and changeable network environments, significantly improving the accuracy and efficiency of detection.

Description

Technical Field

[0001] The present invention relates to the technical field of computer networks, and in particular to a method and system for monitoring a computer network application program test process. Background Art

[0002] Computer network technology covers a wide range of fields and aims to achieve data communication and resource sharing between different devices. Edge gateways are usually located at the edge of the network and play an important role in computer networks. They connect to terminal devices through various communication interfaces (such as Ethernet, Wi-Fi, Bluetooth, ZigBee, RS232 / 485, etc.) and collect data from terminal devices. Edge gateways can distinguish terminal device interfaces (communication protocols and methods), identify and process different types of data, and identify the source of data.

[0003] Before the actual deployment of edge gateways, they are usually tested. Traditional testing methods mainly rely on manual operations and static testing tools to import the communication protocols and data types of terminal devices into virtual machines. However, manual operations cannot meet the dynamic and complex requirements of modern network applications. Existing technologies face many technical problems in the process of network application testing. First, data collection and fusion are not comprehensive enough, and it is difficult to obtain comprehensive information from multiple data sources in real time, resulting in insufficient accuracy and timeliness of anomaly detection. Secondly, most existing anomaly detection algorithms are based on static rules or simple machine learning models, which cannot adapt to complex and changing network environments, and the detection effect is limited. Traditional static models cannot dynamically adapt to changing test environments, resulting in low detection accuracy and efficiency. Summary of the invention

[0004] Based on the above objectives, the present invention provides a method and system for monitoring a computer network application program testing process.

[0005] A method for monitoring a computer network application program test process comprises the following steps:

[0006] S1, Initialize the test environment: Create an isolated test environment based on virtualization technology to ensure high consistency with the actual operating environment and support dynamic adjustment of environmental parameters to meet different test requirements;

[0007] S2, multi-source data fusion: During the test process, multiple data sources are integrated, including network traffic, system logs, user behavior and performance indicators, and real-time data collection and preliminary processing are achieved through edge computing nodes;

[0008] S3, anomaly detection: uses improved deep learning algorithms to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and uses adaptive model update mechanisms to improve detection accuracy and efficiency;

[0009] S4, intelligent feedback: the test results are presented to the tester through a visual tool, and automatic report generation and reminders are supported;

[0010] S5, dynamic resource scheduling: According to the detection results in the intelligent feedback, the computing resources and network resources in the test environment are automatically adjusted to ensure resource utilization and system stability during the test and optimize the test process.

[0011] Furthermore, the initialization test environment specifically includes:

[0012] S11, select the virtualization platform: Select the Hyper-V virtualization platform to create an isolated test environment;

[0013] S12, create a virtual environment template: create an operating system image and an application image that are consistent with the actual operating environment, ensure that the operating system version, software configuration and dependencies are consistent with the actual environment, use Hyper-V Manager to save the operating system image and application image as an image template to quickly deploy virtual machines, configure the network environment, including network parameters such as IP address, subnet mask, gateway, ensure that the network settings are consistent with the actual environment, install and configure software and services, and ensure that the application can run normally in the virtual environment;

[0014] S13, deploy virtual environment: use Hyper-V manager and pre-created image templates to deploy virtual machines. Create the required number of virtual machines according to test requirements, ensure that the configuration of each instance matches the actual environment, allocate computing resources (CPU, memory, storage) to each virtual machine, ensure that the resource configuration is consistent with the actual environment, and use Hyper-V's dynamic memory management function to automatically adjust the memory allocation of virtual machines according to requirements;

[0015] S14, Network isolation: Create an independent virtual network to ensure the isolation of the test environment from the external environment and prevent confusion between test data and actual data. Use the Hyper-V Virtual Switch Manager to configure the virtual network, create internal, external, or dedicated virtual switches, and configure firewall rules and access control policies to ensure that only authorized testers and test tools can access the virtual environment. Use the Hyper-V snapshot function to create environment snapshots to protect the integrity and security of the virtual environment.

[0016] S15, Dynamically adjust environmental parameters: Use Windows PowerShell and Hyper-V modules to realize automated deployment and management of virtual environments, write PowerShell scripts, dynamically adjust the configuration of virtual machines (such as CPU, memory, storage, network, etc.) to meet different test requirements, monitor the operating status of the virtual environment in real time, use Hyper-V performance monitor and resource monitor to collect performance data and resource usage, and automatically or manually adjust environmental parameters based on monitoring data to optimize resource utilization and system performance;

[0017] S16, Environment Verification: Regularly perform consistency checks on the virtual environment to ensure that the environment configuration is consistent with the actual operating environment. Use the configuration management tool Chef in combination with PowerShell DSC (Desired State Configuration) to achieve consistency management and automated repair of the environment.

[0018] Furthermore, in the test process, S2 integrates multiple data sources, including:

[0019] S21, Network traffic data: Use Hyper-V's built-in network monitoring function to capture network traffic data, configure monitoring policies, and ensure that the captured data covers all network interfaces and protocols used for testing;

[0020] S22, system log data: Use Windows Event Log to collect operating system and application logs, configure the collection scope of event logs, including system logs, security logs, and application logs, and store log data in a centralized storage location for subsequent processing and analysis;

[0021] S23, User behavior data: Deploy user behavior analysis tools to collect user operation and interaction data in the application. The collected data includes user clicks, page visits, and session duration;

[0022] S24, performance indicator data: Use the Prometheus performance monitoring tool to monitor system resource usage and application performance, and configure monitoring items, including CPU usage, memory usage, disk I / O, and network latency.

[0023] Furthermore, the real-time data collection and preliminary processing implemented by the edge computing node in S2 specifically include:

[0024] S25, deployment of edge computing nodes: In the test environment, select physical or virtual nodes close to the data source as edge computing nodes, install and configure the AWS Greengrass edge computing platform, and deploy data collection agents on each data source. The data collection agents are responsible for collecting and transmitting data to the edge computing nodes in real time;

[0025] Use encryption technology to protect data transmission and ensure that the communication between data collection agents and edge computing nodes is secure and reliable;

[0026] S26, real-time data collection and preliminary processing: configure the data collection module on the edge computing node to receive data from the data source, use the HTTP protocol adapter, unify the data transmission protocol, and ensure seamless integration of different types of data;

[0027] Implement data filtering and preprocessing functions on edge computing nodes to remove redundant and irrelevant data, reduce data volume, and use the Apache Kafka Streams stream processing framework to preprocess real-time data, including data cleaning, format conversion, and aggregation calculations;

[0028] Temporarily store the processed data in the SQLite local database of the edge computing node for quick access and analysis, and forward the data to the central storage and cloud database regularly or on demand for persistent storage;

[0029] S27, Data Monitoring and Management: Configure a real-time monitoring panel, use the Grafana visualization tool to display the real-time status and trend of data, set alarm rules, and issue alarm notifications in a timely manner when monitoring indicators exceed preset thresholds;

[0030] Regularly synchronize the data on the edge computing nodes to the central storage platform to ensure data consistency and integrity, configure data backup strategies, prevent data loss, and ensure data security during the testing process.

[0031] Furthermore, the S3 uses an improved deep learning algorithm to analyze real-time data and identify potential abnormal behaviors and performance bottlenecks, including:

[0032] S31, data preprocessing: normalize the real-time data to ensure that the data is in the same dimension and eliminate the impact of dimension differences on model training. Use sliding window technology to divide the real-time data stream into several time windows and extract the data features in each window.

[0033] S32, Feature Extraction: Use convolutional neural networks to extract local features of data and long short-term memory networks to capture the temporal dependency of data;

[0034] S33, model training: input the extracted features into the deep neural network for classification and regression analysis, identify normal behavior and abnormal behavior, use an improved loss function, add a penalty term for abnormal behavior, and improve the sensitivity and recognition ability of the deep neural network model to abnormal behavior;

[0035] S34, Model Optimization: Use adaptive learning rate optimization algorithm to train deep neural network models to improve training efficiency and model convergence speed;

[0036] S35, real-time data analysis: the data collected by the edge computing nodes is pre-processed and then input into the deep neural network model in real time for online analysis;

[0037] S36: Abnormal behavior identification: The deep neural network model classifies and regresses the input data to identify abnormal behaviors and performance bottlenecks. By setting an abnormal threshold, when the abnormal score output by the model exceeds the threshold, it is marked as abnormal behavior;

[0038] S37, performance bottleneck detection: perform regression analysis on performance indicator data to identify the bottleneck points of system performance.

[0039] Furthermore, the adaptive model updating mechanism is used in S3 to improve the accuracy and efficiency of detection, specifically including:

[0040] Online learning: Use real-time data streams to incrementally train the deep neural network model, gradually update the deep neural network model parameters, and adapt to the dynamic changes of data;

[0041] Model feedback: Feedback anomaly detection results and performance analysis reports to testers, obtain manual annotations and feedback information, use the feedback information to fine-tune the deep neural network model, correct the prediction error of the deep neural network model, and improve the accuracy of detection;

[0042] Regular updates: Regularly retrain the deep neural network model, use the latest data for full training, and update the weights and structure of the deep neural network model;

[0043] Model evaluation: By continuously monitoring the detection accuracy, recall rate and F1 value of the deep neural network model, the performance of the deep neural network model is evaluated. According to the evaluation results, the structure and parameters of the deep neural network model are dynamically adjusted to ensure the efficiency and accuracy of the deep neural network model.

[0044] Furthermore, the intelligent feedback specifically includes:

[0045] S41, Select and configure visualization tools: Use the Kibana visualization tool to provide chart and data display functions, store the anomaly detection result data in the Elasticsearch database, and the Kibana visualization tool reads data from the database;

[0046] S42, Create a Visual Dashboard: Create a dashboard to display different types of detection results, including network traffic anomalies, system log errors, user behavior anomalies, and performance bottlenecks;

[0047] S43, setting alarms and thresholds: configuring alarm rules, and automatically triggering alarms when the detection results exceed the preset thresholds;

[0048] S44, automatic report generation and reminder: Use the report generation function of the Kibana visualization tool to automatically create regular detection reports (such as daily, weekly, and monthly reports), configure the Jenkins automation tool to regularly generate and distribute reports, send reports to testers and managers, configure real-time reminders, and automatically send reminder notifications to relevant personnel when abnormal detection results trigger alarms. Provide multiple reminder methods, including email notifications, SMS reminders, and system pop-ups.

[0049] Furthermore, the dynamic resource scheduling specifically includes:

[0050] S51, Intelligent Feedback Data Analysis: Receives test result data, including performance bottlenecks, abnormal behaviors, and resource utilization, categorizes and prioritizes test results, identifies resources and issues that need to be optimized, and uses machine learning algorithms to conduct in-depth analysis of test results to identify causes of low resource utilization and system bottlenecks;

[0051] S52, resource adjustment strategy formulation: formulate resource adjustment strategies based on the analysis results, including increasing or decreasing computing resources, adjusting network bandwidth and priority, and determining specific parameters and goals for resource adjustment, such as controlling CPU utilization between 60% and 80%, and memory utilization between 70% and 90%;

[0052] S53, Dynamic Resource Allocation: Use the dynamic resource allocation function of virtualization platforms (such as Hyper-V) to automatically adjust the computing resources of virtual machines based on real-time detection results, dynamically adjust CPU and memory allocation, and ensure that virtual machines can respond promptly when resource requirements change;

[0053] S54, Automatically adjust network resources: Use network management tools (such as NetFlow, QoS) to monitor and adjust network bandwidth. Based on the detection results, dynamically adjust network bandwidth allocation to ensure network performance of applications and services.

[0054] Furthermore, in S51, a machine learning algorithm is used to conduct an in-depth analysis of the detection results to identify the causes of low resource utilization and system bottlenecks, specifically including:

[0055] S511, data collection and preprocessing: collect system operation data, including CPU utilization, memory usage, disk I / O, network bandwidth and response time, collect application logs and error logs, record system exceptions and error information, preprocess the collected data, handle missing values ​​and outliers, ensure data quality, and normalize and standardize the data so that different indicators are on the same dimension to facilitate subsequent analysis;

[0056] S512, cluster analysis to identify the cause of low resource utilization includes the following steps:

[0057] Feature selection: Select features related to resource utilization, including CPU utilization, memory usage, disk I / O, and network traffic;

[0058] Clustering algorithm: Use the K-means clustering algorithm to perform cluster analysis on system performance data, identify different resource utilization patterns, calculate the mean and variance of each indicator, and divide the data into K cluster centers;

[0059] K-means algorithm steps:

[0060] Initialize K cluster centers;

[0061] Calculate the distance from each data point to each cluster center and assign it to the nearest cluster center;

[0062] Update the cluster center to the mean of all data points in the current cluster;

[0063] Repeat the above K-means algorithm steps until the cluster center no longer changes or the maximum number of iterations is reached;

[0064] K-means formula: ,in, is the objective function, is the number of clusters, It is clusters, It is The center of the cluster;

[0065] Result analysis: Analyze the clustering results and identify the specific reasons for low resource utilization. If the CPU utilization and memory usage in a cluster are high, but the system response time is slow, it may be a system bottleneck. Identify the characteristics of different cluster centers and find out the patterns and reasons for low resource utilization.

[0066] S513, regression analysis to identify system bottlenecks includes the following steps:

[0067] Feature selection and construction: Select features related to system bottlenecks, including response time, error rate, CPU load, memory usage, disk I / O, and network latency;

[0068] Regression model training: Use linear regression algorithm to build a system performance prediction model, take system performance indicators as feature input, and response time or error rate as target variable to train the regression model;

[0069] The linear regression calculation is: ,in, is the target variable (e.g. response time), is a characteristic variable (such as CPU utilization), is the regression coefficient;

[0070] Model evaluation and selection: Use cross-validation to evaluate model performance, select the optimal regression model, and evaluate the prediction effect of the model through the mean square error. The mean square error is calculated as: ,in, is the true value, is the predicted value, is the sample size;

[0071] Based on the model prediction results, the bottleneck of system performance is identified. If the regression coefficient of a certain feature is large and its value exceeds a certain threshold, the feature may be the cause of the system bottleneck.

[0072] A system for monitoring the test process of a computer network application program, used to implement the above-mentioned method for monitoring the test process of a computer network application program, comprises the following modules:

[0073] Virtualized test environment module: creates an isolated test environment based on virtualization technology to ensure consistency with the actual operating environment and supports dynamic adjustment of environment parameters;

[0074] Multi-source data fusion module: Integrates multiple data sources, including network traffic, system logs, user behavior, and performance indicators, and implements real-time data collection and preliminary processing through edge computing nodes;

[0075] Anomaly detection module: uses an improved deep learning algorithm to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and uses an adaptive model update mechanism to improve detection accuracy and efficiency. The anomaly detection results directly affect the content and suggestions of intelligent feedback;

[0076] Intelligent feedback module: displays the test results to the tester through visual tools, and supports automatic generation of reports and reminders;

[0077] Dynamic resource scheduling module: Automatically adjusts computing resources and network resources in the test environment based on the detection results in intelligent feedback, ensures resource utilization and system stability during the test, and optimizes the test process.

[0078] The present invention integrates multiple data sources, including network traffic, system logs, user behavior, and performance indicators, and uses edge computing nodes to achieve real-time data collection and preliminary processing. This multi-source data fusion method ensures the comprehensiveness and accuracy of the data and provides a reliable basis for anomaly detection. Through the real-time processing capability of edge computing, potential abnormal behaviors and performance bottlenecks can be discovered in a timely manner, improving the real-time and effectiveness of the test process:

[0079] The present invention uses an improved deep learning algorithm to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and continuously optimize the detection model using an adaptive model update mechanism. Through the combination of machine learning and deep learning technology, the system can dynamically adapt to complex and changing network environments, significantly improving the accuracy and efficiency of detection.

[0080] The present invention can automatically adjust the computing resources and network resources in the test environment to ensure resource utilization and system stability during the test. Through the dynamic resource scheduling module, the system can optimize resource allocation in real time, avoid resource waste and system overload, and improve the operating efficiency and stability of the test environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0082] Figure 1 A flowchart of a method for monitoring a computer network application program test process according to an embodiment of the present invention;

[0083] Figure 2 The present invention is a flowchart of a computer network application program testing process monitoring system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0084] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0085] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0086] like Figure 1 As shown, a method for monitoring a computer network application program test process comprises the following steps:

[0087] S1, Initialize the test environment: Create an isolated test environment based on virtualization technology to ensure high consistency with the actual operating environment and support dynamic adjustment of environmental parameters to meet different test requirements. Initializing the test environment provides the basis and guarantee for subsequent data collection and analysis;

[0088] S2, multi-source data fusion: During the test process, multiple data sources are integrated, including network traffic, system logs, user behavior and performance indicators, and real-time data collection and preliminary processing are realized through edge computing nodes. Multi-source data fusion provides comprehensive and accurate data support for anomaly detection;

[0089] S3, anomaly detection: uses improved deep learning algorithms to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and uses adaptive model update mechanisms to improve detection accuracy and efficiency;

[0090] S4, intelligent feedback: the test results are presented to the tester through a visual tool, and automatic report generation and reminders are supported;

[0091] S5, dynamic resource scheduling: According to the detection results in the intelligent feedback, the computing resources and network resources in the test environment are automatically adjusted to ensure resource utilization and system stability during the test and optimize the test process.

[0092] Initializing the test environment specifically includes:

[0093] S11, Select a virtualization platform: Select the Hyper-V virtualization platform to create an isolated test environment. Hyper-V is a powerful virtualization solution provided by Microsoft that can help create and manage a virtual test environment.

[0094] S12, create a virtual environment template: create an operating system image and an application image that are consistent with the actual operating environment, ensure that the operating system version, software configuration and dependencies are consistent with the actual environment, use Hyper-V Manager to save the operating system image and application image as an image template to quickly deploy virtual machines, configure the network environment, including network parameters such as IP address, subnet mask, gateway, ensure that the network settings are consistent with the actual environment, install and configure software and services, and ensure that the application can run normally in the virtual environment;

[0095] S13, deploy virtual environment: use Hyper-V manager and pre-created image templates to deploy virtual machines. Create the required number of virtual machines according to test requirements, ensure that the configuration of each instance matches the actual environment, allocate computing resources (CPU, memory, storage) to each virtual machine, ensure that the resource configuration is consistent with the actual environment, and use Hyper-V's dynamic memory management function to automatically adjust the memory allocation of virtual machines according to requirements;

[0096] S14, Network isolation: Create an independent virtual network to ensure the isolation of the test environment from the external environment and prevent confusion between test data and actual data. Use the Hyper-V Virtual Switch Manager to configure the virtual network, create internal, external, or dedicated virtual switches, and configure firewall rules and access control policies to ensure that only authorized testers and test tools can access the virtual environment. Use the Hyper-V snapshot function to create environment snapshots to protect the integrity and security of the virtual environment.

[0097] S15, Dynamically adjust environmental parameters: Use Windows PowerShell and Hyper-V modules to realize automated deployment and management of virtual environments, write PowerShell scripts, dynamically adjust the configuration of virtual machines (such as CPU, memory, storage, network, etc.) to meet different test requirements, monitor the operating status of the virtual environment in real time, use Hyper-V performance monitor and resource monitor to collect performance data and resource usage, and automatically or manually adjust environmental parameters based on monitoring data to optimize resource utilization and system performance;

[0098] S16, Environment Verification: Regularly check the consistency of the virtual environment to ensure that the environment configuration is consistent with the actual operating environment. Use the configuration management tool Chef in combination with PowerShell DSC (Desired State Configuration) to achieve consistency management and automated repair of the environment.

[0099] Through the above steps, an isolated test environment is created using the Hyper-V virtualization platform to ensure high consistency with the actual operating environment and support dynamic adjustment of environmental parameters to meet different test requirements. This method not only improves the accuracy and efficiency of the test, but also can flexibly adapt to changing test requirements and environmental changes.

[0100] During the testing process, S2 integrates multiple data sources, including:

[0101] S21, Network traffic data: Use Hyper-V's built-in network monitoring function to capture network traffic data, configure monitoring policies, and ensure that the captured data covers all network interfaces and protocols used for testing;

[0102] S22, system log data: Use Windows Event Log to collect operating system and application logs, configure the collection scope of event logs, including system logs, security logs, and application logs, and store log data in a centralized storage location for subsequent processing and analysis;

[0103] S23, User behavior data: Deploy user behavior analysis tools, such as Google Analytics, Matomo, or custom scripts, to collect user operation and interaction data in the application. The collected data includes user clicks, page visits, and session duration;

[0104] S24, performance indicator data: Use the Prometheus performance monitoring tool to monitor system resource usage and application performance, and configure monitoring items, including CPU usage, memory usage, disk I / O, and network latency.

[0105] In S2, real-time data collection and preliminary processing are realized through edge computing nodes, including:

[0106] S25, deployment of edge computing nodes: In the test environment, select physical or virtual nodes close to the data source as edge computing nodes, install and configure the AWS Greengrass edge computing platform, and deploy data collection agents on each data source. The data collection agents are responsible for collecting and transmitting data to the edge computing nodes in real time;

[0107] There are two types of data collection situations of the edge computing node. One is that the edge computing node actively triggers the terminal device to collect data, and the other is that the terminal device feeds back data to the edge computing node. The first type can be achieved through the application program, and the second type can be achieved by modifying the instructions of the virtual machine. This embodiment mainly introduces the second data collection situation.

[0108] The data acquisition module includes a terminal device module, a device management module, a data module and a data processing module;

[0109] The terminal device module is used to obtain the source information of the terminal device connected to the edge gateway and send it to the device management module;

[0110] The device management module determines the registration status of the terminal device according to the source information obtained by the terminal device module, and determines whether the data module for registering and storing the device information needs to be registered according to the registration status; when the terminal device is not registered, a data module is created and the source data information is stored in the data module;

[0111] The data processing module establishes a data acquisition channel for data acquisition according to the data information in the data module. The data processing module loads the corresponding functional module and sends the source information to the device management module to complete the registration of the newly added source information. The data of the terminal device is collected by the data module and then transmitted to the data processing module through the data acquisition channel for processing.

[0112] The data collected in real time by the edge computing node includes source information obtained by the terminal device module, registration information of the device management module, device information stored in the data module, and valid data after passing through the algorithm module.

[0113] The above-mentioned terminal device is the virtual machine in this embodiment. The virtual machine is used to simulate different terminal devices, and an edge gateway is required to collect the data in the virtual machine. One edge gateway can correspond to multiple virtual machines. In this embodiment, in order to be more in line with the data collection of distributed resources in reality, multiple edge gateways are used to collect data.

[0114] The source information obtained by the terminal device includes the communication protocol, device data and data type of the terminal device. The edge gateway receives data information sent from the terminal device. In this implementation, these two types of data can be detected to determine the CPU usage and traffic usage of the edge gateway to view the ability of the edge gateway to collect data;

[0115] Use encryption technology to protect data transmission and ensure that the communication between data collection agents and edge computing nodes is secure and reliable;

[0116] S26, real-time data collection and preliminary processing: configure the data collection module on the edge computing node to receive data from the data source, use the HTTP protocol adapter, unify the data transmission protocol, and ensure seamless integration of different types of data;

[0117] Implement data filtering and preprocessing functions on edge computing nodes to remove redundant and irrelevant data, reduce data volume, and use the Apache Kafka Streams stream processing framework to preprocess real-time data, including data cleaning, format conversion, and aggregation calculations;

[0118] Temporarily store the processed data in the SQLite local database of the edge computing node for quick access and analysis, and forward the data to the central storage and cloud database regularly or on demand for persistent storage;

[0119] S27, Data Monitoring and Management: Configure a real-time monitoring panel, use the Grafana visualization tool to display the real-time status and trend of data, set alarm rules, and issue alarm notifications in a timely manner when monitoring indicators exceed preset thresholds;

[0120] Regularly synchronize the data on the edge computing nodes to the central storage platform to ensure data consistency and integrity, configure data backup strategies, prevent data loss, and ensure data security during the test process;

[0121] Through the above steps, the integration of multi-source data and real-time data collection and preliminary processing of edge computing nodes can be realized during the test process. This method not only improves the real-time and accuracy of data collection, but also reduces the burden on the central storage and analysis platform, and improves the overall test efficiency and effect.

[0122] Abnormal behaviors include the mismatch between the source information of the terminal device and the registration information, the mismatch between the data information collected by the data module and the data information obtained by the terminal device, and the mismatch between the valid data obtained by the data processing module and the data information.

[0123] S3 uses an improved deep learning algorithm to analyze real-time data and identify potential abnormal behaviors and performance bottlenecks, including:

[0124] S31, data preprocessing: normalize the real-time data to ensure that the data is in the same dimension and eliminate the impact of dimension differences on model training. Use sliding window technology to divide the real-time data stream into several time windows and extract the data features in each window.

[0125] Normalization formula: For each data point , normalized, expressed as:

[0126] ,in, is a dataset, and They are the minimum and maximum values ​​of the data set, respectively. Normalization can map the data to the range of [0,1] to eliminate dimensional differences;

[0127] Sliding window technology: divide the real-time data stream into several time windows. Assume that the length of the time window is , then the data in each time window is expressed as: ;

[0128] S32, Feature Extraction: Use convolutional neural networks to extract local features of data, which is suitable for the analysis of network traffic and system log time series data. Use long short-term memory networks to capture the time dependency of data, which is suitable for modeling user behavior and performance indicator sequence data.

[0129] Convolutional Neural Network: The calculation formula of the convolution layer is: ,in, It is The convolution kernels are located at The output at It is The weight of the convolution kernel, is the bias term, is the activation function;

[0130] The calculation formula of the long short-term memory network unit is:

[0131]

[0132] in, It's the Gate of Forgetfulness. is the input gate, is a candidate memory cell, is a memory unit, is the output gate, is the hidden state, and are weights and bias terms;

[0133] S33, model training: input the extracted features into the deep neural network for classification and regression analysis, identify normal behavior and abnormal behavior, use an improved loss function, add a penalty term for abnormal behavior, and improve the sensitivity and recognition ability of the deep neural network model to abnormal behavior;

[0134] Improved loss function: Use weighted cross-extraction loss function and add a penalty term for abnormal behavior, calculated as:

[0135] ;

[0136] in, is the sample size, is the true label, is the predicted probability, and are the weights of normal behavior and abnormal behavior, respectively;

[0137] S34, Model Optimization: Use adaptive learning rate optimization algorithm to train deep neural network models to improve training efficiency and model convergence speed;

[0138] Use an adaptive learning rate optimization algorithm (such as Adam) for model training:

[0139] ;

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] in, and are the first and second moment estimates, and is the decay rate, is the gradient, is the learning rate, are model parameters, is a smoothing term;

[0145] S35, real-time data analysis: the data collected by the edge computing nodes is pre-processed and then input into the deep neural network model in real time for online analysis;

[0146] S36: Abnormal behavior identification: The deep neural network model classifies and regresses the input data to identify abnormal behaviors and performance bottlenecks. By setting an abnormal threshold, when the abnormal score output by the model exceeds the threshold, it is marked as abnormal behavior;

[0147] The model classifies and regresses the input data, identifies abnormal behavior, and predicts probability Output from the model: ,in, and are the weights and biases of the model, is the activation function;

[0148] S37, performance bottleneck detection: perform regression analysis on performance indicator data to identify the bottleneck of system performance;

[0149] Regression analysis of performance indicator data to predict performance bottlenecks: ,in, is the predicted performance indicator value.

[0150] When abnormal behavior is identified, a database query request is sent to the data acquisition module; after receiving the query permission instruction, the original data is output, and the original data includes the source information, registration information, data information, and valid data of the terminal device.

[0151] First, the source information of the terminal device is compared with the registration information and data information respectively, and whether the two types of information are consistent is determined based on the comparison results. If they are consistent, the valid data is compared with the data information, and the abnormal behavior and the module where the abnormal behavior occurs are obtained based on the two comparison results.

[0152] S3 uses an adaptive model update mechanism to improve detection accuracy and efficiency, including:

[0153] Online learning: Use real-time data streams to incrementally train the deep neural network model, gradually update the deep neural network model parameters, and adapt to the dynamic changes of data;

[0154] Use the incremental learning method to continuously update the model weights according to the new data. Suppose the new data sample is , then the weight update formula is expressed as: ;

[0155] Model feedback: Feedback anomaly detection results and performance analysis reports to testers, obtain manual annotations and feedback information, use the feedback information to fine-tune the deep neural network model, correct the prediction error of the deep neural network model, and improve the accuracy of detection;

[0156] Regular updates: Regularly retrain the deep neural network model, use the latest data for full training, and update the weights and structure of the deep neural network model;

[0157] Model evaluation: Evaluate the performance of the deep neural network model by continuously monitoring the detection accuracy, recall rate and F1 value of the deep neural network model. Based on the evaluation results, dynamically adjust the structure and parameters of the deep neural network model to ensure the efficiency and accuracy of the deep neural network model.

[0158] Accuracy , recall rate , ;

[0159] Through the above steps, the improved deep learning algorithm can be used to analyze real-time data, identify abnormal behaviors and performance bottlenecks, and use the adaptive model update mechanism to improve the accuracy and efficiency of detection. This method not only improves the accuracy and real-time performance of anomaly detection, but also can adapt to the ever-changing test environment and data characteristics through continuous model optimization.

[0160] Smart feedback includes:

[0161] S41, Select and configure visualization tools: Use the Kibana visualization tool to provide chart and data display functions, store the anomaly detection result data in the Elasticsearch database, and the Kibana visualization tool reads data from the database;

[0162] S42, Create a Visual Dashboard: Create a dashboard to display different types of detection results, including network traffic anomalies, system log errors, user behavior anomalies, and performance bottlenecks;

[0163] S43, setting alarms and thresholds: configuring alarm rules, and automatically triggering alarms when the detection results exceed the preset thresholds;

[0164] S44, automatic report generation and reminder: Use the report generation function of the Kibana visualization tool to automatically create regular detection reports (such as daily, weekly, and monthly reports), configure the Jenkins automation tool to regularly generate and distribute reports, and send reports to testers and managers. Configure real-time reminders. When abnormal detection results trigger alarms, reminder notifications are automatically sent to relevant personnel. Provide multiple reminder methods, including email notifications, SMS reminders, and system pop-ups;

[0165] Through the above steps, the test results can be displayed to the tester through the visualization tool, and it also supports automatic generation of reports and reminders, which improves the test efficiency and effectiveness.

[0166] Dynamic resource scheduling specifically includes:

[0167] S51, Intelligent Feedback Data Analysis: Receives test result data, including performance bottlenecks, abnormal behaviors, and resource utilization, categorizes and prioritizes test results, identifies resources and issues that need to be optimized, and uses machine learning algorithms to conduct in-depth analysis of test results to identify causes of low resource utilization and system bottlenecks;

[0168] S52, resource adjustment strategy formulation: formulate resource adjustment strategies based on the analysis results, including increasing or decreasing computing resources, adjusting network bandwidth and priority, and determining specific parameters and goals for resource adjustment, such as controlling CPU utilization between 60% and 80%, and memory utilization between 70% and 90%;

[0169] S53, Dynamic Resource Allocation: Use the dynamic resource allocation function of virtualization platforms (such as Hyper-V) to automatically adjust the computing resources of virtual machines based on real-time detection results, dynamically adjust CPU and memory allocation, and ensure that virtual machines can respond promptly when resource requirements change;

[0170] S54, Automatically adjust network resources: Use network management tools (such as NetFlow, QoS) to monitor and adjust network bandwidth. Based on the detection results, dynamically adjust network bandwidth allocation to ensure network performance of applications and services.

[0171] In S51, machine learning algorithms are used to conduct in-depth analysis of the detection results to identify the causes of low resource utilization and system bottlenecks, including:

[0172] S511, data collection and preprocessing: collect system operation data, including CPU utilization, memory usage, disk I / O, network bandwidth and response time, collect application logs and error logs, record system exceptions and error information, preprocess the collected data, handle missing values ​​and outliers, ensure data quality, and normalize and standardize the data so that different indicators are on the same dimension to facilitate subsequent analysis;

[0173] S512, cluster analysis to identify the cause of low resource utilization includes the following steps:

[0174] Feature selection: Select features related to resource utilization, including CPU utilization, memory usage, disk I / O, and network traffic;

[0175] Clustering algorithm: Use the K-means clustering algorithm to perform cluster analysis on system performance data, identify different resource utilization patterns, calculate the mean and variance of each indicator, and divide the data into K cluster centers;

[0176] K-means algorithm steps:

[0177] Initialize K cluster centers (select randomly or use K-means++);

[0178] Calculate the distance from each data point to each cluster center and assign it to the nearest cluster center;

[0179] Update the cluster center to the mean of all data points in the current cluster;

[0180] Repeat the above steps until the cluster center no longer changes or the maximum number of iterations is reached;

[0181] K-means formula: ,in, is the objective function, is the number of clusters, It is clusters, It is The center of the cluster;

[0182] Result analysis: Analyze the clustering results and identify the specific reasons for low resource utilization. If the CPU utilization and memory usage in a cluster are high, but the system response time is slow, it may be a system bottleneck. Identify the characteristics of different cluster centers and find out the patterns and reasons for low resource utilization.

[0183] S513, regression analysis to identify system bottlenecks includes the following steps:

[0184] Feature selection and construction: Select features related to system bottlenecks, including response time, error rate, CPU load, memory usage, disk I / O, and network latency;

[0185] Regression model training: Use linear regression algorithm to build a system performance prediction model, take system performance indicators as feature input, and response time or error rate as target variable to train the regression model;

[0186] The linear regression calculation is: ,in, is the target variable (e.g. response time), is a characteristic variable (such as CPU utilization), is the regression coefficient;

[0187] Model evaluation and selection: Use cross-validation to evaluate model performance, select the optimal regression model, and evaluate the prediction effect of the model through the mean square error. The mean square error is calculated as: ,in, is the true value, is the predicted value, is the sample size;

[0188] Based on the model prediction results, the bottleneck of system performance is identified. If the regression coefficient of a certain feature is large and its value exceeds a certain threshold, the feature may be the cause of the system bottleneck.

[0189] like Figure 2 As shown, a system for monitoring the test process of a computer network application is used to implement the above-mentioned method for monitoring the test process of a computer network application, and includes the following modules:

[0190] Virtualized test environment module: creates an isolated test environment based on virtualization technology to ensure consistency with the actual operating environment and supports dynamic adjustment of environment parameters;

[0191] Multi-source data fusion module: Integrates multiple data sources, including network traffic, system logs, user behavior, and performance indicators, and implements real-time data collection and preliminary processing through edge computing nodes;

[0192] Anomaly detection module: uses an improved deep learning algorithm to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and uses an adaptive model update mechanism to improve detection accuracy and efficiency. The anomaly detection results directly affect the content and suggestions of intelligent feedback;

[0193] Intelligent feedback module: displays the test results to the tester through visual tools, and supports automatic generation of reports and reminders;

[0194] Dynamic resource scheduling module: Automatically adjusts computing resources and network resources in the test environment based on the detection results in intelligent feedback, ensures resource utilization and system stability during the test, and optimizes the test process.

[0195] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0196] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for monitoring a computer network application program test process, characterized in that: The following steps are involved: S1, Initialize the test environment: Create an isolated test environment based on virtualization technology to ensure high consistency with the actual operating environment and support dynamic adjustment of environment parameters; S2, multi-source data fusion: During the test process, multiple data sources are integrated, including network traffic, system logs, user behavior and performance indicators, and real-time data collection and preliminary processing are achieved through edge computing nodes; S3, anomaly detection: uses improved deep learning algorithms to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and uses adaptive model update mechanisms to detect them; The use of an improved deep learning algorithm to analyze real-time data and identify potential abnormal behaviors and performance bottlenecks specifically includes: S31, data preprocessing: normalize the real-time data to ensure that the data is in the same dimension, use the sliding window technology to divide the real-time data stream into several time windows, and extract the data features in each window; S32, Feature Extraction: Use convolutional neural networks to extract local features of data and long short-term memory networks to capture the temporal dependency of data; S33, model training: the extracted features are input into the deep neural network for classification and regression analysis to identify normal and abnormal behaviors, using an improved loss function and adding a penalty term for abnormal behaviors; S34, Model Optimization: Use adaptive learning rate optimization algorithm to train deep neural network models; S35, real-time data analysis: the data collected by the edge computing nodes is pre-processed and then input into the deep neural network model in real time for online analysis; S36: Abnormal behavior identification: The deep neural network model classifies and regresses the input data to identify abnormal behaviors and performance bottlenecks. By setting an abnormal threshold, when the abnormal score output by the model exceeds the threshold, it is marked as abnormal behavior; S37, performance bottleneck detection: perform regression analysis on performance indicator data to identify the bottleneck of system performance; S4, intelligent feedback: the test results are presented to the tester through a visual tool, and automatic report generation and reminders are supported; S5, dynamic resource scheduling: automatically adjust the computing resources and network resources in the test environment based on the detection results in the intelligent feedback.

2. The method for monitoring the computer network application testing process according to claim 1, characterized in that: The initialization test environment specifically includes: S11, select the virtualization platform: Select the Hyper-V virtualization platform to create an isolated test environment; S12, create a virtual environment template: create an operating system image and an application image that are consistent with the actual operating environment; use the Hyper-V manager to save the operating system image and the application image as an image template; configure the network environment, including the IP address, subnet mask, and gateway; S13, deploy virtual environment: use Hyper-V Manager and pre-created image templates to deploy virtual machines, create the required number of virtual machines according to test requirements, and allocate computing resources to each virtual machine; use Hyper-V's dynamic memory management function to automatically adjust the memory allocation of virtual machines according to requirements; S14, Network Isolation: Create independent virtual networks, use Hyper-V Virtual Switch Manager to configure virtual networks, create internal, external, or dedicated virtual switches, and configure firewall rules and access control policies, and use Hyper-V's snapshot function to create environment snapshots to protect the integrity and security of the virtual environment; S15, Dynamically adjust environmental parameters: Use Windows PowerShell and Hyper-V modules to automate deployment and management of virtual environments, write PowerShell scripts, dynamically adjust virtual machine configurations to meet different test requirements, monitor the operating status of virtual environments in real time, use Hyper-V performance monitors and resource monitors to collect performance data and resource usage, and automatically or manually adjust environmental parameters based on monitoring data; S16, Environment Verification: Regularly perform consistency checks on the virtual environment; use the configuration management tool Chef in combination with PowerShell DSC to achieve consistency management and automated repair of the environment.

3. A method for monitoring a computer network application program testing process according to claim 1, characterized in that: During the testing process, S2 integrates multiple data sources, including: S21, Network traffic data: Use Hyper-V's built-in network monitoring function to capture network traffic data, configure monitoring policies, and ensure that the captured data covers all network interfaces and protocols used for testing; S22, system log data: Use Windows Event Log to collect operating system and application logs, configure the collection scope of event logs, including system logs, security logs, and application logs, and store log data in a centralized storage location; S23, User behavior data: Deploy user behavior analysis tools to collect user operation and interaction data in the application. The collected data includes user clicks, page visits, and session duration; S24, performance indicator data: Use the Prometheus performance monitoring tool to monitor system resource usage and application performance, and configure monitoring items, including CPU usage, memory usage, disk I / O, and network latency.

4. The method for monitoring the computer network application program testing process according to claim 3, characterized in that: The real-time data collection and preliminary processing implemented by the edge computing node in S2 specifically include: S25, deployment of edge computing nodes: In the test environment, select physical or virtual nodes close to the data source as edge computing nodes, install and configure the AWS Greengrass edge computing platform, and deploy data collection agents on each data source. The data collection agents are responsible for collecting and transmitting data to the edge computing nodes in real time; Use encryption technology to protect data transmission; S26, real-time data collection and preliminary processing: configure the data collection module on the edge computing node, receive data from the data source, use the HTTP protocol adapter, and unify the data transmission protocol; Implement data filtering and preprocessing functions on edge computing nodes to remove redundant and irrelevant data, reduce data volume, and use the Apache Kafka Streams stream processing framework to preprocess real-time data, including data cleaning, format conversion, and aggregation calculations; The processed data is temporarily stored in the SQLite local database of the edge computing node; S27, Data Monitoring and Management: Configure a real-time monitoring panel, use the Grafana visualization tool to display the real-time status and trend of data, set alarm rules, and issue alarm notifications in a timely manner when monitoring indicators exceed preset thresholds; Regularly synchronize data on edge computing nodes to the central storage platform.

5. The method for monitoring the computer network application testing process according to claim 1, characterized in that: The detection in S3 using the adaptive model update mechanism specifically includes: Online learning: Use real-time data streams to incrementally train the deep neural network model, gradually update the deep neural network model parameters, and adapt to the dynamic changes of data; Model feedback: Feedback anomaly detection results and performance analysis reports to testers, obtain manual annotations and feedback information, use the feedback information to fine-tune the deep neural network model, and correct the prediction error of the deep neural network model; Regular updates: Regularly retrain the deep neural network model, use the latest data for full training, and update the weights and structure of the deep neural network model; Model evaluation: The performance of the deep neural network model is evaluated by continuously monitoring the detection accuracy, recall rate, and F1 value of the deep neural network model. The structure and parameters of the deep neural network model are dynamically adjusted based on the evaluation results.

6. The method for monitoring the computer network application program testing process according to claim 5, characterized in that: The intelligent feedback specifically includes: S41, Select and configure visualization tools: Use the Kibana visualization tool to provide chart and data display functions, store the anomaly detection result data in the Elasticsearch database, and the Kibana visualization tool reads data from the database; S42, Create a Visual Dashboard: Create a dashboard to display different types of detection results, including network traffic anomalies, system log errors, user behavior anomalies, and performance bottlenecks; S43, setting alarms and thresholds: configuring alarm rules, and automatically triggering alarms when the detection results exceed the preset thresholds; S44, automatic report generation and reminder: Use the report generation function of the Kibana visualization tool to automatically create regular detection reports, configure the Jenkins automation tool to regularly generate and distribute reports, send reports to testers and managers, configure real-time reminders, and automatically send reminder notifications to relevant personnel when abnormal detection results trigger alarms. Provide multiple reminder methods, including email notifications, SMS reminders, and system pop-ups.

7. The method for monitoring the computer network application program testing process according to claim 6, characterized in that: The dynamic resource scheduling specifically includes: S51, Intelligent Feedback Data Analysis: Receives test result data, including performance bottlenecks, abnormal behaviors, and resource utilization, categorizes and prioritizes test results, identifies resources and issues that need to be optimized, and uses machine learning algorithms to conduct in-depth analysis of test results to identify causes of low resource utilization and system bottlenecks; S52, resource adjustment strategy formulation: formulate resource adjustment strategies based on the analysis results, including increasing or decreasing computing resources, adjusting network bandwidth and priority, and determining specific parameters and goals for resource adjustment; S53, dynamic resource allocation: using the dynamic resource allocation function of the virtualization platform, automatically adjust the computing resources of the virtual machine according to the real-time detection results, and dynamically adjust the CPU and memory allocation; S54, automatically adjust network resources: use network management tools to monitor and adjust network bandwidth, and dynamically adjust network bandwidth allocation based on detection results.

8. The method for monitoring the computer network application program testing process according to claim 7, characterized in that: In S51, a machine learning algorithm is used to perform an in-depth analysis on the detection results to identify the causes of low resource utilization and system bottlenecks, specifically including: S511, data collection and preprocessing: collect system operation data, including CPU utilization, memory usage, disk I / O, network bandwidth and response time, collect application logs and error logs, record system exceptions and error information, preprocess the collected data, handle missing values ​​and outliers, and normalize and standardize the data so that different indicators are on the same dimension; S512, cluster analysis to identify the cause of low resource utilization includes the following steps: Feature selection: Select features related to resource utilization, including CPU utilization, memory usage, disk I / O, and network traffic; Clustering algorithm: Use the K-means clustering algorithm to perform cluster analysis on system performance data, identify different resource utilization patterns, calculate the mean and variance of each indicator, and divide the data into K cluster centers; K-means algorithm steps: Initialize K cluster centers; Calculate the distance from each data point to each cluster center and assign it to the nearest cluster center; Update the cluster center to the mean of all data points in the current cluster; Repeat the K-means algorithm steps until the cluster center no longer changes or the maximum number of iterations is reached; K-means formula: Among them, J is the objective function, K is the number of clusters, C i is the i-th cluster, μ i is the center of the i-th cluster; Result analysis: Analyze the clustering results, identify the specific reasons for low resource utilization, identify the characteristics of different cluster centers, and find out the patterns and reasons for low resource utilization; S513, regression analysis to identify system bottlenecks includes the following steps: Feature selection and construction: Select features related to system bottlenecks, including response time, error rate, CPU load, memory usage, disk I / O, and network latency; Regression model training: Use linear regression algorithm to build a system performance prediction model, take system performance indicators as feature input, and response time or error rate as target variable to train the regression model; Linear regression calculation is: y = β0 + β1x1 + β2x2 + ... + β n x n , where y is the target variable, x i is the characteristic variable, β i is the regression coefficient; Model evaluation and selection: Use cross-validation to evaluate model performance, select the optimal regression model, and evaluate the prediction effect of the model through the mean square error. The mean square error is calculated as: Among them, y i is the true value, is the predicted value, n is the number of samples; Based on the model prediction results, identify the bottleneck points of system performance.

9. A computer network application testing process monitoring system, characterized in that: include: The computer network application test process monitoring system is used to implement the computer network application test process monitoring method according to any one of claims 1 to 8; Virtualized test environment module, which creates an isolated test environment based on virtualization technology, ensures consistency with the actual operating environment, and supports dynamic adjustment of environment parameters; Multi-source data fusion module integrates multiple data sources, including network traffic, system logs, user behavior and performance indicators, and realizes real-time data collection and preliminary processing through edge computing nodes; Anomaly detection module, which includes a feature extraction module, a model training and optimization module, a real-time data analysis module, an abnormal behavior identification module, and a performance bottleneck detection module; an improved deep learning algorithm is used to analyze real-time data, identify potential abnormal behaviors and performance bottlenecks, and an adaptive model update mechanism is used for detection. The abnormal detection results directly affect the content and suggestions of intelligent feedback; The feature extraction module uses a convolutional neural network to extract local features of the data and a long short-term memory network to capture the time dependency of the data; The model training and optimization module inputs the extracted features into a deep neural network for classification and regression analysis, identifies normal and abnormal behaviors, uses an improved loss function, and adds a penalty term for abnormal behaviors; and uses an adaptive learning rate optimization algorithm to train the deep neural network model; The real-time data analysis module: pre-processes the data collected by the edge computing node and inputs it into the deep neural network model in real time for online analysis; The abnormal behavior identification module: the deep neural network model classifies and regresses the input data, identifies abnormal behaviors and performance bottlenecks, and sets an abnormal threshold. When the abnormal score output by the model exceeds the threshold, it is marked as abnormal behavior; The performance bottleneck detection module: performs regression analysis on the performance indicator data to identify the bottleneck point of system performance; Intelligent feedback module, which displays the test results to the tester through visual tools and supports automatic generation of reports and reminders; The dynamic resource scheduling module automatically adjusts the computing resources and network resources in the test environment according to the detection results in the intelligent feedback.

Images