Testing method and device based on virtual testing network, equipment and medium

Abstract

The invention relates to the field of network communication, in particular to a test method and device based on a virtual test network, equipment and a medium. According to the embodiment of the invention, the pre-deployed virtual test network is called by acquiring the test instruction, and the historical test data is loaded in the virtual host, so that a foundation is laid for comprehensive test. Historical data are analyzed to obtain a test strategy, the virtual host is tested according to the strategy in the virtual network, original flow can be obtained, basic rate testing is covered, and key indexes such as delay and packet loss probability can be comprehensively captured through subsequent analysis. And the protocol analysis layer analyzes the original traffic to obtain service features, and further deeply mines network performance details. The network performance problem can be accurately found through abnormal detection of service features, the virtual network is optimized and adjusted based on the detection result, and the network performance can be improved in a targeted mode. The series of processes can meet the requirements of the SoC array server board card network performance comprehensive test in a complex network environment, and provide more comprehensive and accurate network performance evaluation.

Description

Technical Field

[0001] The present invention relates to the field of network communications, and in particular to a testing method, device, equipment and medium based on a virtual test network. Background Art

[0002] In the field of server testing, traditional technical solutions for testing network cards in SoC array servers have numerous limitations. Current testing methods typically require at least two servers to separately handle data transmission and reception, significantly increasing hardware costs and leading to low testing efficiency and significant waste of resources.

[0003] Furthermore, in a single-switch environment, all boards and cards share the same network path, making independent testing impossible. This significantly limits the flexibility and specificity of testing. More critically, existing testing methods are limited, supporting only basic rate testing and failing to comprehensively cover metrics critical to server network performance evaluation, such as latency and packet loss rate. This inability to meet the demand for comprehensive network performance testing of SoC array server boards and cards in complex network environments necessitates innovative technical solutions. Summary of the Invention

[0004] In view of this, the embodiments of the present invention provide a testing method, apparatus, equipment and medium based on a virtual test network to solve the problem that existing testing means are relatively simple and can only support basic rate testing, and are difficult to fully cover indicators such as delay and packet loss rate that are crucial to server network performance evaluation, and cannot meet the needs of comprehensive testing of SoC array server board network performance in complex network environments.

[0005] In a first aspect, an embodiment of the present invention provides a testing method based on a virtual test network, the method comprising:

[0006] Obtaining a test instruction, and obtaining a pre-deployed virtual test network according to the test instruction, and loading historical test data on a virtual host;

[0007] Analyze the historical test data to obtain a test strategy, and test the virtual host according to the test strategy in the virtual test network to obtain original traffic generated during the test;

[0008] Parsing the original traffic through a protocol parsing layer to obtain service characteristics of the original traffic;

[0009] Anomaly detection is performed on the service characteristics to obtain a detection result, and an optimization adjustment operation is performed on the virtual network corresponding to the virtual host based on the detection result.

[0010] Furthermore, the acquiring of a pre-deployed virtual test network according to the test instruction includes:

[0011] Obtaining test requirements according to the test instructions;

[0012] Deploy a virtual switch on the SoC board and divide it into virtual LANs based on test requirements;

[0013] Based on the deployed virtual switch, inject test traffic into the virtual LAN through the virtual port, and monitor and record the basic performance data of the test traffic in real time;

[0014] Using the VRF function of the virtual switch to divide at least two test channels, and connecting the sending group and the receiving group boards to the corresponding test channels respectively to build an isolated test channel;

[0015] A virtual test network is built based on the constructed isolated test channel and the virtual switch deployed on the SoC board.

[0016] Furthermore, the method further comprises:

[0017] Determining a target isolated test channel requiring refined configuration adjustment based on the virtual test network;

[0018] Collect basic data in full across multiple dimensions based on the target isolation test channel;

[0019] Comparing the basic data with the expected network optimization target to obtain a performance difference index, and using the performance difference index to generate a configuration adjustment plan corresponding to the target isolation test channel;

[0020] A refined configuration adjustment operation is performed on the target isolated test channel in the virtual test network based on the configuration adjustment solution.

[0021] Furthermore, analyzing the historical test data to obtain a test strategy includes:

[0022] Invoke a reinforcement learning decision engine to analyze the historical test data and obtain analysis results;

[0023] Execute parameter generation logic based on the analysis results to generate a dynamic parameter set;

[0024] Determine the test coverage according to the test requirements corresponding to the test instructions;

[0025] The test coverage range is planned using a dynamic parameter set, and the execution data of each test case is corresponding to the test case. The expected results of the test case are set according to the historical test data to obtain the test strategy.

[0026] Furthermore, parsing the original traffic through the protocol parsing layer to obtain the service characteristics of the original traffic includes:

[0027] Decoding the original traffic through the protocol parsing layer to obtain network data in a readable format;

[0028] Performing in-depth analysis of different protocols based on the network data in the readable format to obtain key data including business field information;

[0029] Performing business field verification on the sessions in the key data and marking the sessions that fail the verification to obtain abnormal key data;

[0030] Data analysis is performed based on the abnormal key data to obtain business characteristics of the original traffic.

[0031] Furthermore, the performing of anomaly detection on the service feature to obtain a detection result includes:

[0032] Inputting the business features into a pre-trained anomaly detection model;

[0033] The anomaly detection model performs pattern matching and anomaly judgment on the business features according to the learned normal business pattern to obtain a set of abnormal data points;

[0034] Anomaly verification is performed based on the abnormal data point set, abnormal data points that fail the abnormal verification are removed from the abnormal data point set, and the detection result is determined according to the number of abnormal data points remaining in the abnormal data point set.

[0035] Furthermore, the anomaly detection model performs pattern matching and anomaly judgment on the business features according to the learned normal business pattern to obtain a set of abnormal data points, including:

[0036] Performing pattern matching operations on the business features based on the learned normal business patterns through the anomaly detection model to identify data points that do not conform to the normal patterns;

[0037] Detect whether the data points that do not match the normal pattern deviate from the preset data points corresponding to the normal pattern;

[0038] The set of abnormal data points is constructed using data points that deviate from the preset data points corresponding to the normal pattern.

[0039] Furthermore, performing optimization and adjustment operations on the virtual network corresponding to the virtual host based on the detection result includes:

[0040] If the detection result shows that there is no abnormality in the service feature, recording the detection result and using the detection result to optimize the network configuration of the virtual network;

[0041] If the detection result indicates that the service feature is abnormal, an alarm is triggered, and the network parameters of the virtual network are intervened according to the intervention strategy.

[0042] Furthermore, intervening in the network parameters of the virtual network according to the intervention strategy includes:

[0043] Performing short-term trend analysis on historical network indicator data of the virtual network using a sliding window trend analysis method to obtain initial trend changes;

[0044] Predicting network indicators for future periods based on the initial trend change, and calculating an adjustment amount based on the network indicators;

[0045] The network parameters of the virtual network are adjusted according to the adjustment amount.

[0046] In a second aspect, an embodiment of the present invention provides a testing device based on a virtual test network, the device comprising:

[0047] An acquisition module, configured to acquire a test instruction, acquire a pre-deployed virtual test network according to the test instruction, and load historical test data on a virtual host;

[0048] An analysis module, configured to analyze the historical test data to obtain a test strategy, and test the virtual host according to the test strategy in the virtual test network to obtain original traffic generated during the test;

[0049] A parsing module, configured to parse the original traffic through a protocol parsing layer to obtain service characteristics of the original traffic;

[0050] The detection module is used to perform anomaly detection on the service characteristics, obtain a detection result, and perform an optimization and adjustment operation on the virtual network corresponding to the virtual host based on the detection result.

[0051] In a third aspect, an embodiment of the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0052] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method of the first aspect or any corresponding embodiment thereof.

[0053] The embodiment of the present application obtains test instructions to call a pre-deployed virtual test network, and loads historical test data on the virtual host, laying the foundation for comprehensive testing. By analyzing historical data to obtain a test strategy, the virtual host is tested according to the strategy in the virtual network to obtain the original traffic, which not only covers the basic rate test, but also can fully capture key indicators such as delay and packet loss rate through subsequent analysis. The protocol parsing layer parses the original traffic, obtains business characteristics, and further delves into the details of network performance. Anomaly detection business characteristics can accurately detect network performance problems, optimize and adjust the virtual network based on the detection results, and improve network performance in a targeted manner. This series of processes can meet the needs of comprehensive testing of SoC array server board network performance in complex network environments, and provide a more comprehensive and accurate network performance evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 is a flowchart of a testing method based on a virtual test network according to some embodiments of the present invention;

[0056] Figure 2 is a flow chart of another testing method based on a virtual test network according to some embodiments of the present invention;

[0057] Figure 3 is a flowchart of another testing method based on a virtual test network according to some embodiments of the present invention;

[0058] Figure 4 is an example diagram of a test process based on a virtual test network according to some embodiments of the present invention;

[0059] Figure 5 is a structural block diagram of a test device based on a virtual test network according to an embodiment of the present invention;

[0060] Figure 6 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0062] According to an embodiment of the present invention, a testing method, apparatus, device, and medium based on a virtual test network are provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0063] In this embodiment, a testing method based on a virtual testing network is provided. Figure 1 FIG. 1 is a flow chart of a test method based on a virtual test network according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0064] Step S101: obtaining a test instruction, obtaining a pre-deployed virtual test network according to the test instruction, and loading historical test data on a virtual host.

[0065] In an embodiment of the present application, first, when a test instruction is received, the entire test process is started and the relevant test environment and tools are initialized. Subsequently, the historical data accumulated from previous tests are read from a preset database or storage medium. These data cover various parameters, results and other information of past tests. After loading is completed, these historical data are input into the reinforcement learning decision engine. The engine analyzes the characteristics of the historical data based on the preset algorithms and models to determine whether the conditions for generating a test strategy are met. If the conditions are met, the test strategy generation link is entered; if not, it may return to reload the historical data or perform other initialization operations.

[0066] In the embodiment of the present application, a pre-deployed virtual test network is obtained according to the test instruction, such as Figure 2 As shown, the following steps A1-A5 are included:

[0067] Step A1: Obtain test requirements according to test instructions.

[0068] First, the system receives test instructions from users or upper-level applications. These instructions are encapsulated in a standardized format (such as JSON or XML) and contain information such as test objectives (such as performance testing and functional testing), test scenarios (such as high concurrency and low latency), and test time requirements. The system uses the instruction parsing module to perform lexical and syntactic analysis and extract key parameters. For example, from the instruction "perform a high-concurrency network throughput test on the SoC board for 30 minutes," the system parses the test object as the SoC board, the test type as a throughput test, the test duration as 30 minutes, and the business scenario as high concurrency. Finally, the parsed information is integrated into a structured test requirements document to clarify the specific direction and constraints of the test.

[0069] Step A2: deploy a virtual switch on the SoC board according to test requirements and divide the virtual local area network.

[0070] Based on the test requirements, the system invokes a virtualization management tool (such as Open vSwitch or VMware NSX) to deploy a virtual switch within the SoC's operating system kernel or user space. During deployment, parameters such as the number of ports and queue management are configured based on the test scale and resource allocation requirements. After deployment, one or more virtual local area networks (VLANs) (such as VLAN 100) are created using the virtual switch's VLAN functionality.

[0071] In step A3, based on the deployed virtual switch, test traffic is injected into the virtual LAN through the virtual port, and basic performance data of the test traffic is monitored and recorded in real time.

[0072] The system uses traffic generation tools (such as Iperf and Traffic Generator) to inject simulated test traffic into the designated virtual LAN through the virtual port of the virtual switch. The injected traffic can be customized according to the test requirements, including traffic type (TCP, UDP), packet size, sending rate, protocol distribution (HTTP, DNS, etc.). At the same time, the traffic monitoring function is enabled in the virtual switch, and traffic data flowing through the virtual port is collected in real time through technologies such as NetFlow and sFlow. The collected data includes but is not limited to basic performance indicators such as the number of packets, number of bytes, source / destination IP addresses, port numbers, latency, and packet loss rate. These data are transmitted to the monitoring server in real time, stored in a time series database (such as InfluxDB), and dynamically displayed through visualization tools (such as Grafana) so that testers can grasp the running status of the test traffic in real time.

[0073] Step A4: Use the VRF function of the virtual switch to divide at least two test channels, and connect the sending group and receiving group boards to the corresponding test channels respectively to build an isolated test channel.

[0074] With the help of the VRF (virtual routing and forwarding) function of the virtual switch, the system creates at least two independent VRF instances. Each instance is equivalent to a logical router with an independent routing table and forwarding rules. For each VRF instance, the system binds the network interface of the sending group board to one of the VRFs, and the network interface of the receiving group board to the other VRF, thus forming two isolated test channels. During the configuration process, an independent IP address space and routing protocol (such as static routing, OSPF) are allocated to each VRF instance to ensure that traffic between different VRFs cannot be directly interconnected. In this way, the test traffic generated by the sending group board can only be transmitted within the VRF channel to which it belongs, and eventually reaches the receiving group board bound to another VRF channel, achieving complete isolation of the test traffic at the logical level and avoiding interference between different test tasks.

[0075] Step A5: Build a virtual test network based on the constructed isolated test channel and the virtual switch deployed on the SoC board.

[0076] Based on the established isolated test channels and virtual switches, the system further integrates network resources to build a complete virtual test network. By configuring the VXLAN protocol, a logical overlay network is created on top of the physical network, enabling virtual network interconnection across physical locations and simulating a complex wide area network environment. At the same time, combined with the SDN controller, virtual switches and VRF channels are centrally managed and policy configured, dynamically adjusting traffic forwarding paths, bandwidth allocation, and QoS policies. In addition, a reinforcement learning algorithm is introduced to automatically optimize the parameters of the virtual test network (such as VRF routing policies and VXLAN tunnel encapsulation parameters) based on real-time monitored test traffic performance data, forming a dynamically adjustable and highly flexible virtual test network environment to meet diverse testing needs.

[0077] In an embodiment of the present application, the method also includes: determining a target isolated test channel that requires refined configuration adjustment based on a virtual test network; collecting basic data in full in multiple dimensions based on the target isolated test channel; comparing the basic data with the expected network optimization target to obtain a performance difference index, and using the performance difference index to generate a configuration adjustment plan corresponding to the target isolated test channel; and performing refined configuration adjustment operations on the target isolated test channel in the virtual test network based on the configuration adjustment plan.

[0078] In the established virtual test network environment, we first used network traffic monitoring tools to monitor the service traffic of each channel in real time, obtaining data such as traffic volume, direction, and traffic change trends. At the same time, we combined the business requirements documents to clarify the specific network performance requirements (such as latency, bandwidth, and packet loss rate) for different services.

[0079] By comprehensively analyzing traffic monitoring data and business needs, we can identify channels where network performance bottlenecks exist or where business needs are not fully met. We can then accurately determine the target isolation test channel that requires refined configuration adjustments and record detailed information about the channel, such as channel identification, connected device nodes, etc., to provide clear goals for subsequent operations.

[0080] For the identified target isolation test channel, basic data is comprehensively collected from multiple dimensions. On the one hand, network management protocols (such as SNMP) are used to communicate with network devices within the channel to obtain the switch's flow table information, including flow table rule matching conditions (such as source IP, destination IP, port number, etc.), action instructions (such as forwarding, discarding, modifying priority, etc.), and flow table hit counts. On the other hand, professional network monitoring tools are used to monitor the VXLAN tunnel, collecting tunnel encapsulation parameters (such as tunnel identifier, encapsulation format, source and destination IP, etc.), transmission performance indicators (such as delay, jitter, packet loss rate, etc.), and tunnel traffic statistics. In addition, the use of channel-related device resources, such as CPU utilization and memory usage, can also be collected. All collected data is integrated and stored to form a comprehensive and accurate basic data set to provide data support for subsequent performance analysis and configuration adjustments.

[0081] The collected basic data is compared in detail with the expected network optimization goals. Using data analysis algorithms and models, the difference between current network performance and the expected goals is calculated, resulting in a series of performance differential indicators, such as latency differential, throughput shortfall, and packet loss rate excess. These performance differential indicators, combined with factors such as network topology, service traffic characteristics, and device performance limitations, are then used to generate configuration adjustment plans for the target isolation test channel using intelligent algorithms (such as machine learning) or expert experience rules.

[0082] Based on the generated configuration adjustment plan, perform detailed configuration adjustments on the target isolated test channel within the virtual test network. First, connect to the relevant network devices within the channel, such as switches and routers, using a reliable network management platform or automated scripting tool. Then, follow the steps in the adjustment plan to modify the device configurations, such as modifying the switch's flow table rules or adjusting VXLAN tunnel parameters. During this process, ensure that each configuration change is accurate and monitor the network status changes in real time after the configuration changes.

[0083] After completing all configuration adjustments, collect network performance data from the target isolated test channel again and compare it with the baseline data before the adjustments to verify whether the configuration adjustments have achieved the expected optimization results. Also, record the adjusted network status information, including the new configuration parameters and network performance indicators, to provide a reference for subsequent network maintenance and optimization.

[0084] The test requirements are obtained through test instructions, virtual switches are deployed on the SoC board, virtual LANs are divided, and test traffic is injected to monitor and record basic performance data. The VRF function is also used to build isolated test channels, and then a virtual test network is built, which provides a platform for comprehensive testing. On this basis, the target isolated test channel is determined, and basic data is collected in full from multiple dimensions to more accurately obtain network performance details. By comparing the basic data with the expected target, the performance difference index is obtained, and a configuration adjustment plan is generated based on this, and the target channel is fine-tuned for configuration adjustment. This series of operations not only covers basic rate testing, but also comprehensively captures key indicators such as latency and packet loss rate, deeply analyzes network performance issues, and performs targeted optimization. This can meet the comprehensive testing needs of the network performance of SoC array server boards in complex network environments and provide a more accurate and comprehensive network performance evaluation.

[0085] Step S102 : Analyze historical test data to obtain a test strategy, and test the virtual host in the virtual test network according to the test strategy to obtain the original traffic generated during the test.

[0086] In an embodiment of the present application, after the reinforcement learning decision engine determines that the conditions for generating a test strategy are met, a specific test strategy is generated based on dynamic parameters (these parameters are generated by the engine based on historical data analysis) combined with business logic and test objectives, such as determining the scope of test cases, execution order, etc. Subsequently, according to the generated test strategy, the target or component is actually tested on the virtual host. During the test process, various data will be collected, including the original traffic generated during the test process. These traffic data will be used for subsequent protocol parsing and business logic analysis.

[0087] Analyzing historical test data to derive a test strategy includes: invoking a reinforcement learning decision engine to analyze historical test data and obtain analysis results. Executing parameter generation logic based on the analysis results to generate a dynamic parameter set. Determining test coverage based on the test requirements corresponding to the test instructions. Using the dynamic parameter set, planning the execution data for each test case corresponding to the test coverage, and setting expected test case results based on historical test data to derive a test strategy.

[0088] First, historical test data is preprocessed, including data cleaning to remove noise and feature standardization to unify the data scale. This preprocessed data is then fed into the reinforcement learning decision engine. The engine employs a time series analysis algorithm to identify temporal dependencies and long-term patterns in the data. It also employs a clustering algorithm to identify similar groups within the data and extract representative business scenarios. Furthermore, association rule mining analyzes the correlations between different business parameters. During the analysis process, the engine continuously interacts with the environment, evaluating the value of different analysis actions based on a pre-set reward function and gradually optimizing the analysis strategy. Ultimately, it outputs analysis results that include data feature patterns, business scenario classifications, and parameter correlations.

[0089] Based on the analysis results output by the reinforcement learning decision engine, we determine the parameter dimensions that require dynamic adjustment, such as test traffic scale, protocol type ratio, and request frequency. For each parameter dimension, we establish a parameter generation model. For example, for the traffic scale parameter, we use a time series prediction model based on historical traffic distribution and business growth trends to generate the traffic value range for the future test cycle.

[0090] For the protocol type ratio parameter, the weight distribution of different protocols in the test is determined based on the business scenario classification results. A random perturbation factor is also introduced to introduce a certain degree of randomness into the parameter generation process to simulate the uncertainty in a real network environment. Through iterative calculations of the parameter generation model, a dynamic parameter set containing multiple parameter dimensions and their value ranges is ultimately generated.

[0091] After receiving the test instructions, the test objectives and constraints are analyzed, such as whether it is a specialized test targeting a specific business function or a comprehensive performance test. Combined with historical test data analysis results, critical paths and high-risk nodes in the business process are identified and prioritized for testing.

[0092] At the same time, considering testing resource limitations and time constraints, a risk prioritization algorithm is used to prioritize test items. Based on the prioritization results, the scope of business functions, data transaction types, and network scenarios covered by the test is determined. For low-risk areas with a high historical test pass rate, the test coverage can be appropriately narrowed to improve testing efficiency. Ultimately, a clear test coverage definition document is produced.

[0093] For each test case within the test coverage, select an appropriate combination of parameter values from the dynamic parameter set as the execution data for that test case. For example, for an HTTP request processing test case, obtain parameter values such as request frequency, request data format, and number of concurrent users from the dynamic parameter set. For test cases involving complex business processes, it is also necessary to consider the timing and dependencies between parameters to ensure that the execution data simulates real-world business scenarios.

[0094] Then, based on historical test data, set expected results for each test case. For quantitative metrics, such as response time and throughput, determine reasonable threshold ranges based on the distribution of historical data. For qualitative metrics, such as business logic correctness, establish expected behavior patterns based on historical success cases. Integrate test case execution data and expected results to form a complete test strategy document, including the test case execution sequence, resource allocation plan, and exception handling plan.

[0095] Step S103: parse the original traffic through the protocol analysis layer to obtain the service characteristics of the original traffic.

[0096] In an embodiment of the present application, raw traffic is parsed through a protocol parsing layer to obtain service characteristics of the raw traffic, including: decoding the raw traffic through the protocol parsing layer to obtain network data in a readable format; performing in-depth parsing of different protocols based on the readable network data to obtain key data containing service field information; performing service field verification on sessions in the key data, and marking sessions that fail verification to obtain abnormal key data; and performing data analysis based on the abnormal key data to obtain service characteristics of the raw traffic.

[0097] First, the protocol parsing layer receives the raw binary traffic data from the network interface and decodes it layer by layer according to the hierarchical structure of the network protocol stack. Starting from the bit stream of the physical layer, it sequentially parses the frame header of the data link layer (such as MAC address), the IP packet of the network layer (such as source IP, destination IP, protocol type), and the TCP / UDP segment of the transport layer (such as port number and sequence number), and finally extracts the original payload of the application layer. For different application layer protocols (such as HTTP, DNS, MQTT), the corresponding protocol decoder is used to convert the binary payload into text or structured format.

[0098] For example, an HTTP decoder parses the request line (e.g., GET / pathHTTP / 1.1), request header fields (e.g., Content-Type), and request body, converting them into readable key-value pairs. The decoding process handles complex situations such as data segmentation, reassembly, and protocol nesting, ultimately outputting network data in a readable format (e.g., JSON or XML), preserving the complete information of the original traffic.

[0099] Based on the decoded network data in a readable format, the protocol parsing layer classifies it according to the protocol type and applies specific protocol parsing rules. For the HTTP protocol, the business parameters in the URI path (such as order ID, user ID) and the JSON / XML data structure in the request body are further parsed to extract business fields such as product ID, transaction amount, and operation type. For custom protocols (such as financial transaction protocols), the meaning of fields in specific locations (such as transaction code, account information) is parsed according to the protocol specifications. During the parsing process, a protocol context association is established, such as associating the HTTP request with the corresponding response through the session ID to construct a complete business transaction. At the same time, the parsed business fields are standardized, such as converting timestamps in different formats into UTC time and unifying the amount field into a standard currency unit. Finally, a key data set containing business field information and protocol context relationships is generated and stored in a structured manner (such as a relational database or a document database).

[0100] For each session in key data, field-level validation is performed based on predefined business rules. The validation content includes:

[0101] 1) Format verification: Check whether the field conforms to the expected format;

[0102] 2) Value range verification: Verify whether the field value is within the legal range (e.g. transaction amount > 0 and ≤ 1,000,000);

[0103] 3) Logical consistency check: Check the logical relationship between multiple fields (for example, when the order status is "paid", the payment amount must be greater than 0);

[0104] 4) Timing verification: Verify whether the time sequence of business operations is reasonable (such as logging in first and then operating).

[0105] For example, the status code field in an HTTP response is verified to be legal (200-599); for "inventory deduction" operations in e-commerce transactions, the inventory balance is verified to be sufficient. During the verification process, a rules engine is used to dynamically match verification rules and record the verification results. Sessions that fail verification are marked as abnormal and the reason for the failure (such as "amount exceeds limit" or "format error") is attached. Ultimately, abnormal key data is generated, containing markers for both normal and abnormal sessions.

[0106] Finally, the abnormal key data is categorized and organized based on protocol type (e.g., TCP, UDP, HTTP), business function module (e.g., login, transaction, query), and timestamp. Key metrics for each data type, such as mean, median, and standard deviation, are calculated to understand the data's central tendency and dispersion. Furthermore, data mining algorithms, such as association rule mining, are used to identify potential connections between different abnormal key data.

[0107] For example, within HTTP protocol abnormal key data, we analyze the correlation between different status codes and request response times. Through these analytical methods, we can extract the business characteristics of the original traffic in terms of business processes, data interaction patterns, response performance, etc.

[0108] Step S104: perform anomaly detection on the service characteristics to obtain a detection result, and perform optimization and adjustment operations on the virtual network corresponding to the virtual host based on the detection result.

[0109] In the embodiment of the present application, anomaly detection is performed on the service characteristics to obtain a detection result, such as Figure 3 As shown, the following steps B1-B3 are included:

[0110] Step B1: Input the business features into the pre-trained anomaly detection model.

[0111] Specifically, after completing business logic analysis and extracting the service features of the raw traffic, these service feature data are passed as input to a pre-trained LSTM anomaly detection model. This model has the ability to learn normal business patterns and is now ready to use this pre-trained knowledge to perform anomaly detection analysis on the input service features.

[0112] In step B2, the anomaly detection model performs pattern matching and anomaly judgment on the business features according to the learned normal business pattern to obtain a set of abnormal data points.

[0113] Specifically, the anomaly detection model performs pattern matching and anomaly determination on the business features based on the learned normal business pattern to obtain a set of anomaly data points. This includes: performing pattern matching on the business features based on the learned normal business pattern to identify data points that do not conform to the normal pattern; detecting whether the data points that do not conform to the normal pattern deviate from preset data points corresponding to the normal pattern; and constructing a set of anomaly data points using data points that deviate from the preset data points corresponding to the normal pattern.

[0114] First, the processed business feature data is fed into a trained anomaly detection model. Based on previously learned parameters and characteristics of normal business patterns, such as the timing of data interactions in normal business processes, the fluctuation range of key business indicators, and the correlations between fields, the model analyzes the input business feature data point by point.

[0115] In the LSTM model, memory cells and gating mechanisms capture long-term data dependencies and determine whether the current data point conforms to the changing trend of the normal pattern. The model calculates the distance between a data point and a hyperplane constructed based on normal data, assessing its probability of belonging to the normal pattern. By calculating a matching score or distance metric between a data point and the normal pattern, the model identifies data points below a threshold or outside the normal range as inconsistent with the normal pattern and marks them as such.

[0116] For identified data points that do not conform to normal patterns, further comparisons are performed with pre-set data points under normal patterns. Pre-set data points are typically derived from statistical analysis of historical normal business data, including the normal value range of business indicators (such as response time between 50-200ms), the standard format and sequence of data interaction, etc.

[0117] By calculating the degree of deviation between the non-conforming data point and the preset data point in terms of numerical value, format, and timing, for example, calculating the difference between the actual response time value and the preset range boundary, checking whether the data field is missing or formatted incorrectly, etc. If the deviation of the data point in any dimension exceeds the pre-set tolerance threshold (such as the response time deviation exceeds 20% of the upper limit of the normal range), the data point is determined to have deviated from the preset data point corresponding to the normal mode.

[0118] Finally, all data points identified as deviating from the normal pattern are collected and organized, stored in a unified data structure and format, and constructed into a collection of abnormal data points. In addition to the original business feature data, each data point in the collection also includes relevant metadata, such as the business session ID to which the data point belongs, the timestamp of the anomaly detection, the specific dimension of the deviation, and the cause of the deviation.

[0119] For example, in an e-commerce transaction scenario, a collection of abnormal data points might include the order ID of an unusually high transaction amount, the corresponding transaction time, the amount outside the normal range, and an explanation of the deviation. This collection provides a complete and accurate data basis for subsequent exception handling, problem location, and optimization.

[0120] Step B3: perform anomaly verification based on the abnormal data point set, remove abnormal data points that fail the anomaly verification from the abnormal data point set, and determine the detection result according to the number of abnormal data points remaining in the abnormal data point set.

[0121] Specifically, first, for the set of abnormal data points that have been formed, a variety of verification methods are used to perform anomaly verification. For example, by comparing and analyzing data in similar scenarios with historical data, check whether the abnormal data points are within the normal fluctuation range; or use statistical models to calculate the degree to which the data points deviate from the mean to determine whether they are truly abnormal. If an abnormal data point is misjudged as abnormal due to non-substantial abnormal factors such as measurement error and transient interference during the verification process, that is, the abnormal verification fails, the data point will be removed from the set of abnormal data points. After the elimination operation is completed, the number of abnormal data points remaining in the set of abnormal data points is counted. If the remaining number is large and exceeds the pre-set threshold, the detection result is determined to be abnormal; if the remaining number is small and does not reach the threshold, the detection result is determined to be normal.

[0122] In an embodiment of the present application, performing optimization and adjustment operations on the virtual network corresponding to the virtual host based on the detection result includes:

[0123] ① If the test result shows that there is no abnormality in the business characteristics, the test result is recorded and used to optimize the network configuration of the virtual network.

[0124] Specifically, ① if the test result shows that there is no abnormality in the business characteristics, the test result will first be recorded in a log file or database, with detailed information such as the test time, test object, test method and results, to form a complete test record file for subsequent tracing and analysis.

[0125] Next, based on the detection results, the system automatically extracts data such as network performance indicators (such as bandwidth utilization and latency jitter range) and traffic distribution patterns, which are reflected in normal business characteristics. Using machine learning algorithms or preset optimization rules, it conducts in-depth analysis of this data to determine whether there is room for optimization in the current virtual network configuration.

[0126] For example, if the detection finds that the bandwidth utilization is low within a certain period of time, the bandwidth allocation strategy will be appropriately adjusted to dynamically allocate idle bandwidth to other virtual hosts or business modules with higher demand; if it is found that the network delay has optimization potential in a specific business scenario, the network configuration of the virtual network will be targetedly optimized by modifying routing rules, adjusting cache strategies, etc., to improve overall network performance and resource utilization efficiency.

[0127] ② If the detection result shows that the business characteristics are abnormal, an alarm will be triggered and the network parameters of the virtual network will be intervened according to the intervention strategy.

[0128] Specifically, if the detection result shows that there is an abnormality in the business characteristics, the alarm mechanism will be triggered, and an alarm notification will be sent to the operation and maintenance personnel, administrators and relevant business managers through various means such as email, SMS, pop-up windows, etc., detailing the time and location of the abnormality (such as the specific virtual host or business process node), the type of abnormality (such as excessive network latency, abnormal traffic surge) and the possible affected business scope, so that relevant personnel can pay attention and handle it in a timely manner.

[0129] At the same time, the most appropriate intervention plan will be automatically matched based on the pre-set intervention strategy library and the specific circumstances of the abnormality.

[0130] For example, if it is detected that a business delay is abnormal due to network congestion, the bandwidth parameters of the virtual network will be adjusted dynamically according to the intervention strategy, the traffic of some non-critical businesses will be restricted, and the network needs of the core business will be prioritized; if the abnormality is caused by a network configuration error of the virtual host, the relevant network parameters (such as IP address, subnet mask, gateway settings, etc.) will be automatically modified to try to restore the normal operation of the network and reduce the impact of the abnormality on the business.

[0131] Specifically, the network parameters of the virtual network are intervened according to the intervention strategy, including: using a sliding window trend analysis method to perform short-term trend analysis on the historical network indicator data of the virtual network to obtain initial trend changes; predicting network indicators for future periods based on the initial trend changes; and calculating adjustment amounts based on the network indicators. The network parameters of the virtual network are adjusted according to the adjustment amounts.

[0132] When using the sliding window trend analysis method to perform short-term trend analysis on the historical network indicator data of the virtual network, a fixed-size sliding window is first set, which slides sequentially on the time series data. Statistical analysis is performed on the network indicator data (such as delay, throughput, etc.) in each window, and statistical quantities such as mean and variance are calculated to capture the short-term change trend and fluctuation characteristics of the data within a certain time range, thereby obtaining the initial trend change.

[0133] Subsequently, based on the initial trend change, a suitable forecasting algorithm (such as the ARIMA forecasting model mentioned above) is used to forecast the network indicators of future cycles, and the possible value range or specific value of the network indicators in a certain time period in the future is predicted.

[0134] Next, the predicted future network indicators are compared with the preset expected indicators or normal ranges. Based on the difference between the two, the adjustment amount required for the virtual network is calculated through specific calculation rules (such as difference ratio, fixed adjustment coefficient, etc.).

[0135] Finally, based on the calculated adjustment amount, dynamic adjustments are made to the relevant network parameters that affect network indicators in the virtual network (such as packet sending interval, bandwidth allocation ratio, routing strategy, etc.) so that the actual operating status of the virtual network can be optimized in the desired direction and network performance can be improved.

[0136] As an example, consider a virtualized test network using a sliding window trend analysis method to analyze short-term trends in historical network latency data. First, a sliding window size of 10 minutes is set over the past two hours of network latency time series data. The mean and variance of the latency data within each window are calculated. It is found that the mean latency within the last three windows has increased from 50ms to 70ms, and the variance has increased, indicating an initial trend of increasing latency and increasing volatility. Subsequently, an ARIMA forecasting model is used to predict network latency for the next 30 minutes, concluding that the latency value is likely to be between 80-90ms. The predicted results are then compared with the preset normal latency range (30-60ms). By calculating the difference ratio, a 20-30ms delay reduction is determined, and the adjustment amount is calculated. Finally, based on the calculated adjustment amount, the virtual network's packet transmission interval is increased from 100ms to 120ms, and bandwidth allocation is adjusted to prioritize critical service traffic and dynamically optimize routing policies, thereby reducing network latency and improving virtual network performance.

[0137] like Figure 4 As shown, a complete test process is as follows:

[0138] Step 1: Obtain test instructions, requiring a comprehensive assessment of the network performance of the newly launched business system to ensure it can stably carry business traffic. Based on the test instructions, operations personnel clearly define the test requirements as simulating peak business traffic pressure and testing network throughput, latency, and packet loss rate.

[0139] According to the requirements, a virtual switch is deployed on the SoC board through the virtualization management tool, and VLAN 100 is divided as a dedicated virtual LAN for testing.

[0140] Use a virtual port to inject test traffic that simulates peak business traffic into VLAN 100, and use a network monitoring tool to record basic traffic performance data in real time, such as bandwidth utilization and the number of packets per second.

[0141] With the help of the VRF function of the virtual switch, two test channels are divided. The sending group board is connected to the VRF1 channel, and the receiving group board is connected to the VRF2 channel to build isolated test channels.

[0142] Based on the isolated test channel and virtual switch described above, network parameters were further configured to successfully build a virtual test network. Simultaneously, historical data from past network tests of similar business systems was loaded onto the virtual host, including performance indicators and test strategies under different traffic loads.

[0143] Step 2: Input the loaded historical test data into the data analysis system. Using machine learning algorithms, the system conducts in-depth analysis of the data, combining the characteristics of the new business system and the test requirements to generate a test strategy. For example, parameters such as the protocol ratio of the test traffic (e.g., 70% TCP, 30% UDP) and the packet rate of 10,000 packets per second are determined.

[0144] Test the virtual host in a virtual test network according to the generated test strategy. Use professional network testing tools (such as IxChariot) to simulate various network interaction scenarios of the business system, such as user login, data query, file transfer, etc., to generate the original traffic during the test.

[0145] Step 3: The raw traffic generated during the test is transferred to the protocol parsing layer. The protocol parsing layer first decodes the raw traffic, identifying packets of different protocols, such as TCP, UDP, and HTTP. It then performs in-depth analysis of each protocol. For example, for HTTP, it parses out service fields such as the request method (GET, POST, etc.), URL, and HTTP status code; for TCP, it analyzes parameters such as the sequence number, acknowledgment number, and window size. Through these parsing operations, it extracts service characteristics of the raw traffic, such as the frequency of service requests and response time distribution.

[0146] Step 4: Input the extracted service feature data into an anomaly detection model based on the LSTM (Long Short-Term Memory) network. This model has been trained on a large amount of normal and abnormal service traffic data and is able to analyze and judge the current service features.

[0147] If an anomaly is detected, such as a sudden increase in the proportion of HTTP status code 500 error responses or network latency exceeding normal thresholds, the system immediately triggers an alarm and notifies operations and maintenance personnel. Simultaneously, it uses trend prediction modules (e.g., based on sliding window trend analysis and ARIMA prediction models) to predict future trends in network performance indicators, calculate adjustments (such as adjusting packet transmission intervals, optimizing network routing strategies, etc.), and perform optimization adjustments on the virtual network corresponding to the virtual host to improve network performance.

[0148] If no anomalies are detected: The test results, including various performance indicators, are recorded. These results are updated to the historical database for reference in subsequent similar tests, and the test strategy is optimized to more accurately and efficiently evaluate network performance in future tests.

[0149] This embodiment also provides a test device based on a virtual test network, which is used to implement the above-mentioned embodiments and preferred implementations. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0150] This embodiment provides a test device based on a virtual test network, such as Figure 5 Shown, including:

[0151] The acquisition module 501 is configured to acquire a test instruction, acquire a pre-deployed virtual test network according to the test instruction, and load historical test data on a virtual host.

[0152] The analysis module 502 is used to analyze historical test data to obtain a test strategy, and to test the virtual host according to the test strategy in the virtual test network to obtain the original traffic generated during the test.

[0153] The parsing module 503 is configured to parse the original traffic through a protocol parsing layer to obtain service characteristics of the original traffic.

[0154] The detection module 504 is configured to perform anomaly detection on the service characteristics, obtain a detection result, and perform optimization and adjustment operations on the virtual network corresponding to the virtual host based on the detection result.

[0155] In an embodiment of the present application, an acquisition module 501 is used to deploy a virtual switch on the SoC board according to test requirements and divide a virtual LAN; based on the deployed virtual switch, inject test traffic into the virtual LAN through the virtual port, and monitor and record the basic performance data of the test traffic in real time; use the VRF function of the virtual switch to divide at least two test channels, and connect the sending group and receiving group boards to the corresponding test channels respectively to build an isolated test channel; based on the constructed isolated test channel and the virtual switch deployed on the SoC board, build a virtual test network.

[0156] In an embodiment of the present application, the device also includes: an adjustment module, which is used to determine a target isolated test channel that needs to perform fine-grained configuration adjustment based on a virtual test network; collect basic data in full based on multiple dimensions of the target isolated test channel; compare the basic data and the expected network optimization target to obtain a performance difference index, and use the performance difference index to generate a configuration adjustment plan corresponding to the target isolated test channel; and perform fine-grained configuration adjustment operations on the target isolated test channel in the virtual test network based on the configuration adjustment plan.

[0157] In this embodiment of the present application, analysis module 502 is configured to invoke a reinforcement learning decision engine to analyze historical test data and obtain analysis results. Parameter generation logic is executed based on the analysis results to generate a dynamic parameter set. Test coverage is determined based on the test requirements corresponding to the test instructions. The dynamic parameter set is used to plan the execution data for each test case corresponding to the test coverage, and the expected results of the test case are set based on the historical test data to obtain a test strategy.

[0158] In this embodiment of the present application, parsing module 503 is configured to decode the raw traffic through the protocol parsing layer to obtain network data in a readable format. Based on the readable network data, in-depth parsing of different protocols is performed to obtain key data containing business field information. Business fields are verified for sessions in the key data, and sessions that fail verification are marked to obtain abnormal key data. Data analysis is performed based on the abnormal key data to obtain business characteristics of the raw traffic.

[0159] In this embodiment of the present application, detection module 504 is configured to input service features into a pre-trained anomaly detection model. The anomaly detection model then performs pattern matching and anomaly determination on the service features based on the learned normal service patterns, thereby obtaining a set of abnormal data points. Anomaly verification is then performed based on the set of abnormal data points. Abnormal data points that fail the anomaly verification are removed from the set of abnormal data points, and a detection result is determined based on the number of abnormal data points remaining in the set of abnormal data points.

[0160] In this embodiment of the present application, parsing module 503 is configured to perform pattern matching on service features based on the learned normal service pattern using an anomaly detection model, thereby identifying data points that do not conform to the normal pattern. The data points that do not conform to the normal pattern are detected to determine whether they deviate from preset data points corresponding to the normal pattern. A set of anomaly data points is constructed using data points that deviate from the preset data points corresponding to the normal pattern.

[0161] In this embodiment of the present application, the detection module 504 is configured to, if the detection result indicates that the service characteristics are normal, record the detection result and use the detection result to optimize the network configuration of the virtual network. If the detection result indicates that the service characteristics are abnormal, trigger an alarm and intervene in the network parameters of the virtual network according to the intervention policy.

[0162] In this embodiment of the present application, detection module 504 is configured to perform short-term trend analysis on historical network indicator data of the virtual network using a sliding window trend analysis method to obtain an initial trend change. Based on the initial trend change, network indicators for future periods are predicted, and adjustment amounts are calculated based on the network indicators. Network parameters of the virtual network are adjusted based on the adjustment amounts.

[0163] See also Figure 6 , Figure 6 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system).

[0164] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0165] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0166] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created based on the use of a computer device for displaying a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0167] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0168] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0169] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0170] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A testing method based on a virtual test network, characterized in that: The method comprises: Obtaining a test instruction, and obtaining a pre-deployed virtual test network according to the test instruction, and loading historical test data on a virtual host; Analyze the historical test data to obtain a test strategy, and test the virtual host according to the test strategy in the virtual test network to obtain original traffic generated during the test; Parsing the original traffic through a protocol parsing layer to obtain service characteristics of the original traffic; Anomaly detection is performed on the service characteristics to obtain a detection result, and an optimization adjustment operation is performed on the virtual network corresponding to the virtual host based on the detection result.

2. The method according to claim 1, characterized in that The obtaining of a pre-deployed virtual test network according to the test instruction includes: Obtaining test requirements according to the test instructions; Deploy a virtual switch on the SoC board and divide it into virtual LANs based on test requirements; Based on the deployed virtual switch, inject test traffic into the virtual LAN through the virtual port, and monitor and record the basic performance data of the test traffic in real time; Using the VRF function of the virtual switch to divide at least two test channels, and connecting the sending group and the receiving group boards to the corresponding test channels respectively to build an isolated test channel; A virtual test network is built based on the constructed isolated test channel and the virtual switch deployed on the SoC board.

3. The method according to claim 2, characterized in that The method further comprises: Determining a target isolated test channel requiring refined configuration adjustment based on the virtual test network; Collect basic data in full across multiple dimensions based on the target isolation test channel; Comparing the basic data with the expected network optimization target to obtain a performance difference index, and using the performance difference index to generate a configuration adjustment plan corresponding to the target isolation test channel; A refined configuration adjustment operation is performed on the target isolated test channel in the virtual test network based on the configuration adjustment solution.

4. The method according to claim 1, wherein The analyzing the historical test data to obtain a test strategy includes: Invoke a reinforcement learning decision engine to analyze the historical test data and obtain analysis results; Execute parameter generation logic based on the analysis results to generate a dynamic parameter set; Determine the test coverage according to the test requirements corresponding to the test instructions; The test coverage range is planned using a dynamic parameter set, and the execution data of each test case is corresponding to the test case. The expected results of the test case are set according to the historical test data to obtain the test strategy.

5. The method according to claim 1, wherein The parsing of the original traffic through the protocol parsing layer to obtain the service characteristics of the original traffic includes: Decoding the original traffic through the protocol parsing layer to obtain network data in a readable format; Performing in-depth analysis of different protocols based on the network data in the readable format to obtain key data including business field information; Performing business field verification on the sessions in the key data and marking the sessions that fail the verification to obtain abnormal key data; Data analysis is performed based on the abnormal key data to obtain business characteristics of the original traffic.

6. The method according to claim 1, characterized in that The performing anomaly detection on the service feature to obtain a detection result includes: Inputting the business features into a pre-trained anomaly detection model; The anomaly detection model performs pattern matching and anomaly judgment on the business features according to the learned normal business pattern to obtain a set of abnormal data points; Anomaly verification is performed based on the abnormal data point set, abnormal data points that fail the abnormal verification are removed from the abnormal data point set, and the detection result is determined according to the number of abnormal data points remaining in the abnormal data point set.

7. The method according to claim 6, characterized in that The abnormality detection model performs pattern matching and abnormality judgment on the business features according to the learned normal business pattern to obtain a set of abnormal data points, including: Performing pattern matching operations on the business features based on the learned normal business patterns through the anomaly detection model to identify data points that do not conform to the normal patterns; Detect whether the data points that do not match the normal pattern deviate from the preset data points corresponding to the normal pattern; The set of abnormal data points is constructed using data points that deviate from the preset data points corresponding to the normal pattern.

8. The method according to claim 1, characterized in that The performing an optimization and adjustment operation on the virtual network corresponding to the virtual host based on the detection result includes: If the detection result shows that there is no abnormality in the service feature, recording the detection result and using the detection result to optimize the network configuration of the virtual network; If the detection result indicates that the service feature is abnormal, an alarm is triggered, and the network parameters of the virtual network are intervened according to the intervention strategy.

9. The method according to claim 8, characterized in that The intervening in the network parameters of the virtual network according to the intervention strategy includes: Performing short-term trend analysis on historical network indicator data of the virtual network using a sliding window trend analysis method to obtain initial trend changes; Predicting network indicators for future periods based on the initial trend change, and calculating an adjustment amount based on the network indicators; The network parameters of the virtual network are adjusted according to the adjustment amount.

10. A test device based on a virtual test network, characterized in that: The device comprises: An acquisition module, configured to acquire a test instruction, acquire a pre-deployed virtual test network according to the test instruction, and load historical test data on a virtual host; An analysis module, configured to analyze the historical test data to obtain a test strategy, and test the virtual host according to the test strategy in the virtual test network to obtain original traffic generated during the test; A parsing module, configured to parse the original traffic through a protocol parsing layer to obtain service characteristics of the original traffic; The detection module is used to perform anomaly detection on the service characteristics, obtain a detection result, and perform an optimization and adjustment operation on the virtual network corresponding to the virtual host based on the detection result.

Images