Network dial test link generation method and device, network dial test method and device, medium and product

Abstract

The invention discloses a network dial test link generation method and device, a network dial test method and device, a medium and a product. Traffic data of a target network node of a target data center are collected in real time; caching the traffic data to a memory; taking a preset time interval as a training period, and obtaining all the traffic data in the training period from the memory as sample traffic data; and training the sample traffic data by adopting a preset CNN-Kmeans algorithm, and generating a network dial test link for executing a network dial test task. By adopting the method provided by the invention, the flexibility and effectiveness of generating the network dial test link can be improved, so that the network efficiency of a data center can be improved.

Description

Technical Field

[0001] This invention relates to the field of cloud computing technology, and in particular to the generation of network dial-up links, network dial-up methods, devices, media and products. Background Technology

[0002] With the rapid development of cloud computing, users' demand for cloud migration is increasing, and data centers are growing in scale, exhibiting a multi-data center model. As data center scale expands, network virtualization becomes more sophisticated, network topologies become more complex, and the location and localization of network faults become increasingly difficult. How to quickly detect faults has become a pressing problem for technical personnel to solve.

[0003] Existing technologies have proposed generating Pinglists (network test links) for network testing of devices and bare metal objects in data centers, enabling rapid discovery of network faults. However, existing Pinglist generation methods rely on fixed network topologies or manual experience-based summaries. Since network topologies are not static, a Pinglist that may be applicable to one data center may not be well-suited for others, lacking flexibility. Furthermore, existing Pinglist generation methods generate numerous redundant or invalid test tasks, making network problem discovery costly; the trade-off between test coverage and test cost is unsuitable. Summary of the Invention

[0004] The purpose of this invention is to provide a method, device, medium, and product for generating network testing links, which can improve the flexibility and effectiveness of generating network testing links, thereby improving the network efficiency of data centers.

[0005] To achieve the above objectives, embodiments of the present invention provide a method for generating network dial-up test links, comprising: Real-time collection of traffic data from target network nodes in the target data center; The traffic data is cached in memory; A preset time interval is used as the training period. All traffic data within the training period are retrieved from the memory and used as sample traffic data. The sample traffic data is trained using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and generate network dialing links through Kmeans clustering analysis.

[0006] As an improvement to the above scheme, the CNN learning model includes convolutional layers, activation layers, pooling layers, fully connected layers, and normalization layers; The step of training the sample traffic data using a preset CNN-Kmeans algorithm to generate network dialing links includes: The sample traffic data is input into the CNN learning model, and the model sequentially passes through convolutional layers, activation layers, pooling layers, fully connected layers, and normalization layers to perform feature extraction, invalid feature filtering, feature compression, and feature fusion processing on the sample traffic data, and outputs the final feature vector. The K-means algorithm is used to divide the feature vectors into several clusters; Each cluster is considered as a type of business communication scenario. Traffic data corresponding to key feature vectors are selected from each cluster to generate network dialing links.

[0007] As an improvement to the above scheme, the K-means algorithm is used to divide the feature vectors into several clusters, including: A first preset number of feature vectors are selected from the feature vectors as the initial cluster centers; Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; Update the cluster center of each of the aforementioned clusters; Re-execute: Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; The process continues until a preset convergence condition is met, at which point the final cluster is obtained. The convergence condition includes at least one of the following: the change in the cluster center is less than a preset change threshold, a preset number of iterations is reached, and the loss function tends to be minimized.

[0008] As an improvement to the above scheme, after training the sample traffic data using a preset CNN-Kmeans algorithm to generate network dial-up links, the method further includes: The system queries the target data center's basic equipment information, network topology relationships, and network testing targets using a pre-defined configuration management database. Based on the device basic information, the network topology association, and the network testing target, the generated network testing links are corrected to obtain the final network testing links; wherein, the correction process includes at least one of the following: removing redundant network testing links, supplementing missing network testing links, and integrating network testing links according to preset network testing targets.

[0009] As an improvement to the above solution, the real-time acquisition of traffic data of the target network nodes in the target data center includes: Filter and identify the target network node from all network nodes in the target data center; Collect downlink traffic data packets for each of the target network nodes to generate initial traffic data; The initial traffic data of all target network nodes on the target data center is aggregated and analyzed, and invalid traffic data is deleted; wherein, the invalid traffic data includes, but is not limited to, loopback traffic data, known probe traffic data, and attack traffic data; The filtered initial traffic data is sorted in descending order of traffic, and a second preset number of initial traffic data is selected as the final traffic data; wherein, the traffic data includes IP, port and number of traffic bytes.

[0010] This invention also provides a network dialing test method, including: Generate several network testing links corresponding to the target data center; wherein, the network testing links are generated according to the network testing link generation method described in any of the above items; Based on the preset mapping relationship between the test execution unit and the test task, the network test link corresponding to the test task is obtained and sent to the test execution unit for execution; wherein, the test task is set for each service communication scenario, and the test execution unit is deployed on each network node of the target data center.

[0011] This invention also provides a device for generating network dial-up test links, comprising: The data acquisition module is used to collect traffic data of the target network nodes in the target data center in real time. A data caching module is used to cache the traffic data to a memory; The data training module is used to retrieve all traffic data within the training period from the memory as sample traffic data at a preset time interval; and to train the sample traffic data using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and to generate network dialing links through Kmeans algorithm clustering analysis.

[0012] This invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the network dialing link generation method as described in any of the above claims or the network dialing method as described in any of the above claims.

[0013] This invention also provides a computer-readable storage medium, which includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform a network dialing link generation method as described in any one of the preceding claims or a network dialing method as described in any one of the preceding claims.

[0014] This invention also provides a computer program product, which includes a computer program or computer instructions. When the computer program or computer instructions are executed by a processor, they implement the network dialing link generation method or the network dialing method as described in any of the above embodiments.

[0015] Compared with existing technologies, the network testing link generation, network testing method, equipment, media, and products disclosed in this invention collect real-time traffic data from various network nodes in the target data center and perform key feature extraction and cluster analysis using the CNN-Kmeans algorithm. This dynamically generates network testing tasks that comprehensively cover all business communication scenarios within the data center. The real traffic data reflects the actual business communication patterns of the data center, which helps improve the adaptability and accuracy of the generated network testing tasks. CNN extracts features based on the data center's specific traffic data, rather than applying uniform rules. The Kmeans algorithm achieves intelligent clustering of links with similar features, dynamically adapting to topology changes. This improves the flexibility and effectiveness of the generated network testing tasks, reduces the generation of invalid network testing links, saves testing resource overhead, and increases fault detection speed, achieving more accurate fault monitoring and thus improving the network efficiency of the data center. Furthermore, the method for generating network testing tasks proposed in this invention can spontaneously generate network testing links, reducing the difficulty of manual maintenance. It has a high degree of automation and scalability. When updating the data center or adding a new data center resource pool, it can quickly adapt to the network topology of the resource pool without manual configuration. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for generating network dial-up links according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the process for collecting traffic data in an embodiment of the present invention; Figure 3 This is a schematic diagram of the process for analyzing and processing traffic data in an embodiment of the present invention; Figure 4 This is a flowchart illustrating a network dialing test method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a network dial-up link generation device provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a network dialing test device provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0019] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0021] See Figure 1 This is a flowchart illustrating a method for generating a network dial-up test link according to an embodiment of the present invention. The embodiment of the present invention provides a method for generating a network dial-up test link, including steps S11 to S14: S11. Real-time collection of traffic data of target network nodes in the target data center; S12. Cache the traffic data in the memory; S13. Using a preset time interval as the training period, retrieve all the traffic data within the training period from the memory as sample traffic data; S14. The sample traffic data is trained using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and generate network dialing links through Kmeans algorithm cluster analysis.

[0022] In this embodiment of the invention, the data center is equipped with several network nodes, such as several network interface cards of physical servers, bare metal devices, ToR (Top of Rack) switches, etc.

[0023] The network testing link generation method of this invention is applied to a network testing system. This system architecture includes at least a data acquisition module, a data caching module, and a data training module. When a network testing task is required for a target data center, the data acquisition module collects downlink traffic source information (i.e., traffic data) of the target network nodes in the target data center in real time. The collected traffic data is then reported to the data caching module, which caches the traffic data in a memory. This traffic data is used for subsequent model training. The data training module uses a preset time interval as the training period. Within each training period, using all the traffic data collected within that training period as the base data, the CNN-Kmeans algorithm is used to train the model, generating a network testing link plinglist.

[0024] Compared to traditional methods that rely on fixed network topology and manual experience to generate network testing links, this invention uses real-time traffic data from each network node in the target data center. It then employs the CNN-Kmeans algorithm for key feature extraction and clustering analysis to dynamically generate network testing tasks that comprehensively cover all business communication scenarios within the data center. The real traffic data reflects the actual business communication patterns of the data center, improving the adaptability and accuracy of the generated network testing tasks. CNN extracts features based on the data center's unique traffic data, rather than applying uniform rules. The Kmeans algorithm intelligently clusters links with similar features, dynamically adapting to topology changes. This enhances the flexibility and effectiveness of the generated network testing tasks, reduces invalid network testing links, saves testing resource overhead, and improves fault detection speed, enabling more accurate fault monitoring and ultimately improving data center network efficiency. Furthermore, the method for generating network testing tasks proposed in this invention can spontaneously generate network testing links, reducing manual maintenance difficulty. It boasts a high degree of automation and scalability, quickly adapting to the network topology of resource pools when updating the data center or adding new resource pools without manual configuration.

[0025] As a preferred embodiment, the present invention further implements the above embodiments. Step S11, namely the real-time collection of traffic data of the target network node of the target data center, includes steps S111 to S114: S111. Filter and determine the target network node from all network nodes in the target data center; S112. Collect downlink traffic data packets for each target network node to generate initial traffic data; S113. Summarize and analyze the initial traffic data of all target network nodes on the target data center, and delete invalid traffic data; wherein, the invalid traffic data includes, but is not limited to, loopback traffic data, known probe traffic data, and attack traffic data; S114. Sort the filtered initial traffic data in descending order of traffic, and select a second preset number of initial traffic data as the final traffic data; wherein, the traffic data includes IP, port and number of traffic bytes.

[0026] Preferably, step S112 specifically involves: collecting downlink traffic data packets of each target network node through a dialing execution unit deployed on the target network node in the target data center to generate initial traffic data.

[0027] Optionally, the dialing execution unit is a dialing agent (intelligent agent). For example, by building a data collection function into the dialing agent, it is used to periodically analyze the downlink traffic of the current dialing source node using the encapsulated libpcap utility class, obtain the IP and port of the Top N downlink traffic in the current time period, and form traffic data.

[0028] In an embodiment of the present invention, see Figure 2 This is a flowchart illustrating the process of collecting traffic data in this embodiment of the invention. Taking a server network interface card (NIC) as an example, the process begins by obtaining a list of server NICs. For each NIC, traffic from a specified list is excluded, such as traffic from local virtualization NICs, as testing itself is meaningless; only traffic from service NICs is collected. Next, multi-threaded traffic capture and analysis is initiated, outputting downlink traffic information for each NIC. This results in a file named after the NIC, with the file content formatted as: IP / domain name, port, traffic (bytes), sorted in descending order of traffic.

[0029] Further, see Figure 3This is a flowchart illustrating the process of analyzing and processing traffic data in this embodiment of the invention. Multiple traffic files named after network interface cards are merged and sorted in descending order of traffic volume to obtain a summary result. During the summary process, some traffic needs to be excluded based on a specified IP / domain name list, such as loopback traffic, known probe traffic, and identified attack traffic, before the final analysis result is obtained. The top N results are selected according to the service settings, forming a tuple of "source IP, destination IP, destination port, traffic," and reported to the data caching module.

[0030] Optionally, when the dialing agent reports heartbeat data, the data acquisition module reports the traffic data to the data cache module as a serialization result defined by protobuf (structured data serialization and transmission format).

[0031] Based on this, the data caching module provides a corresponding write interface for processing Agent probe results data, Agent heartbeat data, and data reported by the built-in acquisition module.

[0032] For Agent heartbeat messages, if the data portion of the message has a length of 0, it indicates that it is a normal heartbeat message, and only the Agent status refresh time in the backend NoSQL is updated.

[0033] For Agent heartbeat messages, if the data portion of the message is not zero in length and the type in the protocol header is a collection type, the data portion of the message is treated as a serialization result defined by protobuf. In this case, the data caching module deserializes the data portion and writes the result into the Topic storage created by the message queue each day.

[0034] Using the technical means of this invention, an Agent data collection method is implemented to achieve accurate identification of business traffic and noise suppression through analysis and filtering algorithms. The data collection module analyzes and filters the collected data, accurately identifies the traffic attribution, blocks non-business traffic and security-related interference traffic, focuses on business traffic, and reduces noise data caused by collection, providing a good foundation for algorithm training.

[0035] As a preferred embodiment, the present invention is further implemented based on any of the above embodiments, and the CNN learning model includes convolutional layers, activation layers, pooling layers, fully connected layers and normalization layers.

[0036] Then step S14, which is to train the sample traffic data using the preset CNN-Kmeans algorithm to generate network dial-up links, includes steps S141 to S143: S141. Input the sample flow data into the CNN learning model, and sequentially pass through convolutional layers, activation layers, pooling layers, fully connected layers and normalization layers to perform feature extraction, invalid feature filtering, feature compression and feature fusion processing on the sample flow data, and output the final feature vector. S142. The feature vectors are divided into several clusters using the K-means algorithm; S143. Taking each cluster as a type of business communication scenario, select the traffic data corresponding to the key feature vector from each cluster to generate a network dialing link.

[0037] In this embodiment of the invention, the data training module uses the message queue used by the data caching module to synchronize the corresponding daily collected data at fixed time intervals. Based on this data, the CNN-Kmeans algorithm is used to train the collected and analyzed data, and the Pinglist is formed by combining clustering.

[0038] The CNN-Kmeans algorithm is a method that combines Convolutional Neural Networks (CNNs) and the K-means clustering algorithm. A CNN is a deep learning model consisting of multiple layers, primarily including an input layer, convolutional layers, activation layers, pooling layers, fully connected layers, normalization layers, and an output layer. The K-means algorithm is a classic unsupervised clustering algorithm that iteratively divides data points into K clusters, minimizing the sum of the distances between each data point and the centroid of its cluster. Using CNNs effectively extracts features, and the key feature data is then fed into K-means clustering analysis to select the optimal Pinglist.

[0039] In the input layer of a CNN, the key metrics of the collected traffic data, such as traffic volume, destination address, destination port, and source address, are converted into vector representations.

[0040] In a convolutional layer, a set of learnable convolutional kernels (or filters) is used to extract features from the input data to obtain a feature map.

[0041]

[0042] in It is the original input data. It is a convolution kernel. It is the location of the output feature map. It is the index of an element in the convolution kernel.

[0043] In the activation layer, non-linearity enables the network to learn more complex function mappings. The ReLU (Rectified Linear Unit) activation function is used to perform a non-linear transformation on the feature map, filtering out invalid features. The ReLU activation function is as follows:

[0044] In the pooling layer, the feature map is compressed to reduce its spatial size, the number of parameters, and the computational cost. Max pooling is used here to reduce complexity.

[0045] The fully connected layer is located at the end of the network, integrating the features extracted from the previous layers. Each neuron is connected to all activations of the previous layer, and a weighted sum is performed to obtain the fused features.

[0046] in, It's weight. It is input. It is a bias.

[0047] The normalization layer performs normalization operations to improve the network's generalization ability. The softmax loss function is used to calculate the feature probability distribution, outputting the final feature vector. The difference between the predicted values ​​and the true labels is calculated as follows:

[0048] in, It represents the total number of categories. It is the c-th element in the one-hot encoding of the real label. It is 1 if the sample belongs to category c, and 0 otherwise. It is the probability that the model predicts that the sample belongs to class c.

[0049] Further, step S142, namely, dividing the feature vector into several clusters using the K-means algorithm, includes: A first preset number of feature vectors are selected from the feature vectors as the initial cluster centers; Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; Update the cluster center of each of the aforementioned clusters; Re-execute: Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; The process continues until a preset convergence condition is met, at which point the final cluster is obtained. The convergence condition includes at least one of the following: the change in the cluster center is less than a preset change threshold, a preset number of iterations is reached, and the loss function tends to be minimized.

[0050] In this embodiment of the invention, the K-means algorithm is an unsupervised learning algorithm used to divide data points into K clusters. The goal of the algorithm is to minimize the variance within a cluster or to minimize the sum of the distances from all data points to their cluster centers. The specific steps are as follows: The feature vectors obtained from CNN training are fed into K-means initialization. The K-means algorithm starts by selecting K data points as initial cluster centers (centroids). These initial cluster centers can be selected randomly or using a specific heuristic method.

[0051] At the allocation layer, for each data point, the algorithm calculates its distance to each cluster center and assigns it to the nearest cluster center. Euclidean distance is typically used to measure this distance.

[0052] Where x is a data point. It is the first The center of each cluster, This represents the Euclidean distance.

[0053] Next, the cluster centers need to be updated. Once all data points have been assigned to clusters, the center of each cluster needs to be updated. The cluster center is the mean of all its members:

[0054] in, It is the set of all data points assigned to cluster i. It is a set The number of data points in the middle.

[0055] The means algorithm attempts to minimize the loss function that is the sum of the squared distances from all data points to their cluster centers, known as the Within-Cluster Sum of Squares (WCSS).

[0056] The cluster centers are repeatedly assigned and updated until any of the following conditions are met: the change in cluster centers is less than a preset threshold, the preset maximum number of iterations is reached, or the loss function WCSS no longer decreases significantly. The iteration stops when the algorithm meets the convergence condition. The cluster center and data point assignments at this point are considered the final result.

[0057] Finally, each cluster corresponds to a type of business traffic characteristics with common business communication scenarios. The top n traffic data with the highest traffic share and the widest coverage are selected from each cluster to form several network dialing links.

[0058] This invention proposes a method for generating a high-coverage Pinglist by combining CNN and K-means algorithms, which reduces invalid probe links and improves data center network efficiency. The CNN-K-means algorithm generates a comprehensive Pinglist, effectively reducing invalid probe links in the data center network and improving overall network efficiency.

[0059] As a preferred embodiment, the present invention further implements the above embodiments. In step S14, that is, after training the sample traffic data using a preset CNN-Kmeans algorithm to generate the network dial-up link, the method further includes steps S15 and S16: S15. Query the basic equipment information, network topology relationships, and network testing targets of the target data center through the preset configuration management database; S16. Based on the device basic information, the network topology association, and the network testing target, the generated network testing link is corrected to obtain the final network testing link; wherein, the correction process includes at least one of the following: removing redundant network testing links, supplementing missing network testing links, and integrating network testing links according to a preset network testing target.

[0060] In this embodiment of the invention, the network probing system architecture further includes a data correction module. The data correction module is combined with an external CMDB (Configuration Management Database) system to optimize the Pinglist generated during training, mainly by removing redundant probing links.

[0061] The CMDB system serves as a device and topology dictionary for the data center, pre-managing all relevant network resources and providing query interfaces. It provides key data such as basic device information, network topology relationships, manually configured network testing targets (including manually configured network testing links), etc.

[0062] In this embodiment of the invention, the Pinglist results generated by the data training model also need to be modified in conjunction with the CMDB, including removing redundant test links; further trimming some terminal test targets connected to PODs or TORs, and appropriately increasing the test of TOR switches to improve test coverage; and integrating network test links based on the manually configured network test links stored in the CMDB.

[0063] As an example, removing redundant probes: When querying the CMDB, it is found that for a certain probe target, such as server IP1, there are multiple probe sources. If IP2 and IP3 are two probe sources under TOR1, then one probe source needs to be removed.

[0064] Increase TOR probe coverage: When querying the CMDB, if for a certain probe target, such as the server IP4, if the network devices connected to IP4, such as TOR2 or even POD1, are not set as probe targets, then increase the probe links from the IP4 probe source to TOR2 and POD1 to improve probe coverage.

[0065] Integrated dialing test: Querying the CMDB revealed that the original Pinglist was manually configured. By integrating the trained Pinglist with the original Pinglist and removing duplicate entries, a valid network dialing test link was obtained.

[0066] Preferably, the method further includes: The final network test links after correction are grouped and stored according to the test tasks; wherein, the test tasks are set up for each service communication scenario.

[0067] By employing the technical methods of this invention, the entire network of devices is managed through a CMDB (Content Management Database). Information on all network devices is obtained via a query interface, and redundancy is removed and supplemented in network testing links to effectively improve coverage. Network testing links are grouped by task, and the Pinglist is dynamically updated, overcoming the shortcomings of previous Pinglist configurations based on manual experience. This significantly reduces the performance loss caused by the central service calculating the Pinglist through network topology and the additional overhead caused by the agent collecting invalid data.

[0068] See Figure 4 This is a flowchart illustrating a network dialing test method provided in an embodiment of the present invention. The present invention also provides a network dialing test method, including steps S21 to S22: S21. Generate several network testing links corresponding to the target data center; wherein, the network testing links are generated according to the network testing link generation method described in any of the above embodiments; S22. Based on the preset mapping relationship between the test execution unit and the test task, obtain the network test link corresponding to the test task and send it to the test execution unit for execution; wherein, the test task is set for each service communication scenario, and the test execution unit is deployed on each network node of the target data center.

[0069] In this embodiment of the invention, the network dialing test system architecture further includes a task distribution module. The task distribution module is responsible for distributing the generated Pinglist results and updating the existing dialing test tasks. After receiving the updated results, the dialing test Agent updates the dialing test target to achieve the purpose of dynamically adjusting the dialing test scope.

[0070] Specifically, the testing agents are predefined to handle which testing tasks, and a mapping relationship between testing agents and testing tasks is generated. For example, an agent deployed on a regular server in a rack carries rack-based testing tasks and testing tasks for devices connected to the same ToR switch, and is responsible for probing other servers in the same rack and the same ToR switch; an agent deployed on a core server in a data center carries cross-rack testing tasks and cross-data center testing tasks, and is responsible for probing core devices in other racks and other data centers; an agent deployed in a POD container carries container-based service testing tasks and POD-connected device testing tasks, and is responsible for probing services within the same POD and the POD gateway to which it belongs.

[0071] The task distribution module reads the final Pinglist from the database, serializes it in protobuf format, generates update messages, and retrieves the latest Pinglist information for each testing task based on the mapping relationship between the testing agent and its assigned testing tasks. The testing agent receives the update messages from the task distribution module and performs protobuf deserialization to obtain the Pinglist list and testing parameters for each testing task. At the start of the next testing task, the testing agent uses the latest list and other testing parameters to complete the task. Understandably, the data acquisition module in the testing agent records round-trip time (RTT), packet loss rate, and other data, and reports the probe data in protobuf format to the data cache module for subsequent model training and optimization.

[0072] By employing the technical means of this invention, the task distribution module establishes a connection with the testing agent and distributes each agent's exclusive Pinglist separately. This reduces the burden on the agents, eliminating the need for them to handle testing tasks unrelated to themselves, thus reducing CPU and network overhead on the servers where the agents reside. Furthermore, the task distribution module only distributes the necessary links, reducing network transmission volume and improving distribution efficiency. If adjustments to a certain type of testing task are required subsequently, only the update needs to be distributed to the agent carrying that task, without needing a full push, making operations and maintenance more flexible.

[0073] See Figure 5 This is a schematic diagram of a network dial-up link generation device provided in an embodiment of the present invention. The embodiment of the present invention provides a network dial-up link generation device 10, comprising: Data acquisition module 11 is used to collect traffic data of target network nodes in the target data center in real time; Data caching module 12 is used to cache the traffic data to a memory; The data training module 13 is used to retrieve all traffic data within the training period from the memory as sample traffic data at a preset time interval; and to train the sample traffic data using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and to generate network dialing links through Kmeans algorithm clustering analysis.

[0074] In a preferred embodiment, the device 10 further includes a data correction module 14, which is used to query the device basic information and network topology association of the target data center through a preset configuration management database; and to correct the generated network testing links according to the device basic information and the network topology association to obtain the final network testing links; wherein the correction process includes at least one of the following: removing redundant network testing links, supplementing missing network testing links, and integrating network testing links according to preset testing targets.

[0075] It should be noted that the network dialing link generation device provided in this embodiment of the invention is used to execute all the process steps of the network dialing link generation method of the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0076] See Figure 6 This is a schematic diagram of the structure of a network dialing test device provided in an embodiment of the present invention. The present invention also provides a network dialing test device 20, comprising: The network test link generation module 21 is used to generate a number of network test links corresponding to the target data center; wherein, the network test links are generated according to the network test link generation method described in any of the above embodiments; The task distribution module 22 is used to obtain the network testing link corresponding to the testing task according to the preset mapping relationship between the testing execution unit and the testing task, and distribute it to the testing execution unit for execution; wherein, the testing task is set for each business communication scenario, and the testing execution unit is deployed on each network node of the target data center.

[0077] It should be noted that the network dialing test device provided in this embodiment of the invention is used to execute all the process steps of the network dialing test method in the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0078] This invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the network dialing link generation method as described in any of the above embodiments or the network dialing method as described in any of the above embodiments.

[0079] This invention also provides a computer-readable storage medium, which includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform a network dialing link generation method as described in any of the above embodiments or a network dialing method as described in any of the above embodiments.

[0080] This invention also provides a computer program product, which includes a computer program or computer instructions. When the computer program or computer instructions are executed by a processor, they implement the network dialing link generation method or the network dialing method as described in any of the above embodiments.

[0081] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0082] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for generating a network dial-up link, characterized in that, include: Real-time collection of traffic data from target network nodes in the target data center; The traffic data is cached in memory; A preset time interval is used as the training period. All traffic data within the training period are retrieved from the memory and used as sample traffic data. The sample traffic data is trained using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and generate network dialing links through Kmeans clustering analysis.

2. The method for generating network dial-up test links as described in claim 1, characterized in that, The CNN learning model includes convolutional layers, activation layers, pooling layers, fully connected layers, and normalization layers; The step of training the sample traffic data using a preset CNN-Kmeans algorithm to generate network dialing links includes: The sample traffic data is input into the CNN learning model, and the model sequentially passes through convolutional layers, activation layers, pooling layers, fully connected layers, and normalization layers to perform feature extraction, invalid feature filtering, feature compression, and feature fusion processing on the sample traffic data, and outputs the final feature vector. The K-means algorithm is used to divide the feature vectors into several clusters; Each cluster is considered as a type of business communication scenario. Traffic data corresponding to key feature vectors are selected from each cluster to generate network dialing links.

3. The method for generating network dial-up links as described in claim 2, characterized in that, The K-means algorithm is used to divide the feature vectors into several clusters, including: A first preset number of feature vectors are selected from the feature vectors as the initial cluster centers; Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; Update the cluster center of each of the aforementioned clusters; Re-execute: Calculate the Euclidean distance between each feature vector and the cluster center, and assign the feature vector to the nearest cluster based on the Euclidean distance; The process continues until a preset convergence condition is met, at which point the final cluster is obtained. The convergence condition includes at least one of the following: the change in the cluster center is less than a preset change threshold, a preset number of iterations is reached, and the loss function tends to be minimized.

4. The method for generating network dial-up test links as described in claim 1, characterized in that, After training the sample traffic data using a preset CNN-Kmeans algorithm to generate network dial-up links, the method further includes: The system queries the target data center's basic equipment information, network topology relationships, and network testing targets using a pre-defined configuration management database. Based on the device basic information, the network topology association, and the network testing target, the generated network testing links are corrected to obtain the final network testing links; wherein, the correction process includes at least one of the following: removing redundant network testing links, supplementing missing network testing links, and integrating network testing links according to preset network testing targets.

5. The method for generating network dial-up test links as described in claim 1, characterized in that, The real-time acquisition of traffic data from target network nodes in the target data center includes: Filter and identify the target network node from all network nodes in the target data center; Collect downlink traffic data packets for each of the target network nodes to generate initial traffic data; The initial traffic data of all target network nodes on the target data center is aggregated and analyzed, and invalid traffic data is deleted; wherein, the invalid traffic data includes, but is not limited to, loopback traffic data, known probe traffic data, and attack traffic data; The filtered initial traffic data is sorted in descending order of traffic, and a second preset number of initial traffic data is selected as the final traffic data; wherein, the traffic data includes IP, port and number of traffic bytes.

6. A network dialing test method, characterized in that, include: Generate several network testing links corresponding to the target data center; wherein, the network testing links are generated according to the network testing link generation method as described in any one of claims 1 to 5; Based on the preset mapping relationship between the test execution unit and the test task, the network test link corresponding to the test task is obtained and sent to the test execution unit for execution; wherein, the test task is set for each service communication scenario, and the test execution unit is deployed on each network node of the target data center.

7. A device for generating network dial-up link test links, characterized in that, include: The data acquisition module is used to collect traffic data of the target network nodes in the target data center in real time. A data caching module is used to cache the traffic data to a memory; The data training module is used to retrieve all the traffic data within the training period from the memory at a preset time interval as sample traffic data. The sample traffic data is trained using a preset CNN-Kmeans algorithm to generate network dialing links; wherein, the CNN-Kmeans algorithm is used to extract features of the traffic data through a CNN learning model and generate network dialing links through Kmeans clustering analysis.

8. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for generating a network dial-up link as described in any one of claims 1 to 5 or the network dial-up method as described in claim 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the network dialing link generation method as described in any one of claims 1 to 5 or the network dialing method as described in claim 6.

10. A computer program product, characterized in that, The computer program product includes a computer program or computer instructions, which, when executed by a processor, implement the method for generating a network dial-up test link as described in any one of claims 1 to 5 or the network dial-up test method as described in claim 6.

Images