A method, device and medium for locating network faults
By generating undirected graphs and using graph attention network algorithms to model communication networks, the problems of excessively high node feature dimensions and insufficient generalization ability are solved, achieving efficient and accurate fault location and intelligent management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INSPUR NETWORK TECH (SHANDONG) CO LTD
- Filing Date
- 2024-07-22
- Publication Date
- 2026-05-26
AI Technical Summary
Existing technologies represent all port information on node features, resulting in excessively high dimensionality of node features and insufficient generalization ability of the model to port connection changes, making it difficult to quickly and accurately locate communication network faults.
By identifying communication devices with related relationships, an undirected graph is generated, and a graph attention network algorithm is used to model and classify the edges of the undirected graph to identify network fault points.
It improves the accuracy of fault location, reduces the workload of network maintenance personnel, enhances the stability and reliability of communication networks, and realizes intelligent management of communication networks.
Smart Images
Figure CN118802502B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of fault location technology, and in particular to a network fault location method, device and medium. Background Technology
[0002] Currently, fault location is a crucial step in the routing and switching network environment, directly impacting network stability and service continuity. However, traditional fault location methods primarily rely on manual troubleshooting by network engineers. This approach is not only time-consuming and labor-intensive but also often struggles to quickly and accurately pinpoint the root cause of faults in complex and ever-changing network environments. With the rapid development of deep learning technology, especially the significant advantages of Graph Neural Networks (GNNs) in processing graph data, applying them to fault location in communication networks has become a new research hotspot. GNNs can capture the complex relationships within graph structures and identify abnormal patterns in the network by learning node features and the connections between nodes, thus promising automatic fault location.
[0003] Existing inventions attempt to map network devices as graph nodes, encode key device information as graph features, and utilize Generative Neural Networks (GNNs) for modeling and fault root cause identification. However, since communication network devices typically have multiple ports, and not all ports participate in network connections simultaneously, representing all port information in node features leads to excessively high node feature dimensionality. This not only increases computational complexity but also introduces a large amount of redundant information. This unused port information does not help with fault localization but instead interferes with the model's judgment. Furthermore, since port connections in the network may change frequently, fixing all port information in node features results in poor model generalization ability to port connection changes. Summary of the Invention
[0004] This application provides a network fault location method, device, and medium to solve the technical problem that the prior art represents all port information on node features, resulting in excessively high node feature dimensionality and insufficient generalization ability of the model to port connection changes.
[0005] On the one hand, embodiments of this application provide a network fault location method, including:
[0006] Based on the actual communication network, identify every two associated communication devices; wherein, the associated communication devices refer to neighboring nodes that have signal transmission requirements;
[0007] The port states of two associated communication devices are concatenated to obtain the corresponding edge features, and an undirected graph corresponding to the actual communication network is generated based on the edge features.
[0008] Based on the graph attention network algorithm, the undirected graph is modeled and the edges in the undirected graph model are classified to achieve fault location of the actual communication network.
[0009] In one implementation of this application, determining each pair of associated communication devices based on the actual communication network specifically includes:
[0010] Obtain the actual communication network with fault location requirements, and determine all signal transmission directions in the actual communication network;
[0011] Based on each signal transmission direction, the neighboring nodes with signal transmission needs are identified, and the corresponding signal senders and receivers are determined.
[0012] Based on the signal sender and the signal receiver, each pair of associated communication devices in the actual communication network is identified.
[0013] In one implementation of this application, the step of concatenating the port states of two associated communication devices to obtain corresponding edge features specifically includes:
[0014] Determine the port status corresponding to the communication device, and obtain a preset number of port status values from the two associated communication devices respectively; wherein, the port status includes: the port's packet transmit / receive count, maximum transmission unit configuration, rate configuration, MAC address, IP address, as well as device operating temperature, system resource consumption, and on / off status;
[0015] The port status values of the two associated communication devices are randomly concatenated to obtain the edge features corresponding to the edges between the two associated communication devices.
[0016] In one implementation of this application, the modeling of the undirected graph based on the graph attention network algorithm specifically includes:
[0017] Based on the undirected graph, multiple neighbor edge features corresponding to each edge feature are determined; wherein, the edge connected to the corresponding nodes of the communication devices at both ends of the edge in the undirected graph is regarded as the neighbor edge of the edge.
[0018] Based on the graph attention network algorithm, the edge features and the neighboring edge features corresponding to the edge features are aggregated and calculated to obtain the next layer edge features corresponding to the edge features.
[0019] After the edge features are aggregated in multiple layers, the node classification representations corresponding to the aggregated edge features are output through a fully connected layer.
[0020] In one implementation of this application, the graph attention network algorithm is used to aggregate the edge features and the neighboring edge features corresponding to the edge features to obtain the next-layer edge features corresponding to the edge features, specifically including:
[0021] The edge features of the next layer are calculated using the following formula:
[0022]
[0023] in, Let N represent the edge feature of the i-th edge in the next layer, σ represent the activation function, K represent the number of aggregation heads in this layer, and N represent the number of aggregation heads in this layer. i W represents the characteristics of all neighboring edges of the i-th edge. k Let represent the linear transformation matrix of the edge features, and be the parameters to be learned. This represents the importance coefficient between the i-th edge and the features of each neighboring edge.
[0024] In one implementation of this application, the graph attention network algorithm is used to aggregate the edge features and the neighboring edge features corresponding to the edge features to obtain the next-layer edge features corresponding to the edge features, specifically including:
[0025] The importance coefficient between an edge and the features of each of its neighboring edges is calculated using the following formula:
[0026]
[0027] Where || denotes the vector concatenation operation, A k Represents the vector to be learned. This represents performing a linear transformation on the l-th layer feature vectors of i and j and then concatenating them, A k Length and The lengths are the same. A represents k and When performing the inner product, LeakyReLU() represents the activation function, and exp() represents the exponential function.
[0028] In one implementation of this application, classifying the edges in the undirected graph model to locate faults in the actual communication network specifically includes:
[0029] Based on a pre-trained graph attention network model, features are extracted from each edge in the undirected graph model, and the extracted features are classified and judged.
[0030] The fault edges that are classified as faults are identified, and based on the edge features of the fault edges, the two associated communication devices corresponding to the fault edges are determined.
[0031] A fault location report is generated for the two associated communication devices, and the fault location report is sent to the corresponding management personnel to achieve fault location of the actual communication network.
[0032] In one implementation of this application, the step of extracting features from each edge of the undirected graph model based on a pre-trained graph attention network model and classifying the extracted features specifically includes:
[0033] Based on a pre-trained graph attention network model, in an undirected graph model, the packet count, maximum transmission unit configuration, rate configuration, MAC address, IP address, device operating temperature, system resource consumption, and on / off status of the port corresponding to each edge are extracted.
[0034] The edge features are input into the pre-trained graph attention network model to perform forward propagation calculation on the edge features and output the failure probability of each edge as a failure category.
[0035] The relationship between the fault probability and a preset threshold is determined, and based on this relationship, it is determined whether each edge belongs to a fault category.
[0036] On the other hand, embodiments of this application also provide a network fault location device, the device comprising:
[0037] At least one processor;
[0038] And, a memory communicatively connected to the at least one processor;
[0039] The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform a network fault location method as described above.
[0040] On the other hand, embodiments of this application also provide a non-volatile computer storage medium storing computer-executable instructions, which, when executed, implement a network fault location method as described above.
[0041] This application provides a network fault location method, device, and medium, which includes at least the following:
[0042] Beneficial effects:
[0043] By identifying interconnected communication devices and generating an undirected graph based on their port states, the topology of the actual communication network and the relationships between devices can be more accurately reflected. Using graph attention network algorithms to model and classify edges in the undirected graph enables more precise identification of fault points in the network, thus improving the accuracy of fault location. Automating the identification and analysis of relationships between communication devices and fault location based on these relationships significantly reduces the workload of network maintenance personnel, making communication network maintenance more efficient and systematic, and contributing to improved overall maintainability. Modeling and analyzing communication networks using graph attention network algorithms enables intelligent management of the communication network. Accurate fault location facilitates rapid response and repair of network problems, thereby reducing the impact of network failures on communication services. Timely detection and handling of potential fault points can prevent larger-scale network failures, thus improving the stability and reliability of the communication network. Attached Figure Description
[0044] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0045] Figure 1 A flowchart illustrating a network fault location method provided in an embodiment of this application;
[0046] Figure 2 A schematic diagram of an undirected graph constructed based on a network, provided as an embodiment of this application;
[0047] Figure 3 This is a schematic diagram of the edge features of an undirected graph provided in an embodiment of this application;
[0048] Figure 4 This is a schematic diagram of the internal structure of a network fault location device provided in an embodiment of this application. Detailed Implementation
[0049] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0050] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.
[0051] Figure 1This is a flowchart illustrating a network fault location method provided in an embodiment of this application.
[0052] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.
[0053] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.
[0054] like Figure 1 As shown in the embodiment of this application, a network fault location method includes:
[0055] 101. Based on the actual communication network, identify every two associated communication devices; where associated communication devices refer to neighboring nodes that have signal transmission requirements.
[0056] Specifically, in one embodiment of this application, the server obtains the actual communication network with fault location requirements and determines all signal transmission directions in the actual communication network; based on each signal transmission direction, it determines the neighboring nodes with signal transmission requirements and determines the corresponding signal sender and signal receiver; based on the signal sender and signal receiver, it determines every two related communication devices in the actual communication network.
[0057] In one embodiment, an actual communication network with fault location requirements is first obtained. This network includes multiple switches, routers, and terminal devices. Then, through network management tools or protocols (such as SNMP, LLDP, or wireless SNMP, NetFlow, etc.), device information in the network and the connection relationships between them are collected to determine the direction of all signal transmissions.
[0058] For each signal transmission direction, identify the neighboring nodes with signal transmission needs. For example, if a signal is transmitted from switch A to switch B, then switch A and switch B are neighboring nodes. Determine the signal sender and receiver for each transmission direction. In the example above, switch A is the signal sender, and switch B is the signal receiver.
[0059] Based on the information from the signal sender and receiver, each pair of related communication devices in the actual communication network is identified. For example, if switch A has signal transmission relationships with switch B, router C, and terminal device D, then these devices are all related to switch A and can be considered as related communication devices.
[0060] 102. Concatenate the port states of two related communication devices to obtain the corresponding edge features, and generate an undirected graph corresponding to the actual communication network based on the edge features.
[0061] Specifically, in one embodiment of this application, the server determines the port status corresponding to the communication device and obtains a preset number of port status values from the two associated communication devices. It should be noted that the port status in this embodiment includes: the port's packet transmit / receive count, Maximum Transmission Unit (MTU) configuration, rate configuration, MAC address, IP address, device operating temperature, system resource consumption, and on / off status. The specific port status values retrieved can be customized according to actual conditions. The port's MAC address or IP address needs to be normalized; the normalized MAC address or IP address can be used as an edge feature.
[0062] Then, the port status values of the two associated communication devices are randomly concatenated to obtain the edge features corresponding to the edges between the two associated communication devices.
[0063] In one embodiment, in a real communication network, two associated communication devices are selected, such as switch A and switch B. For switch A and switch B, their port status information is obtained using a network management tool or protocol (such as SNMP). Then, a preset number is set, for example, selecting 5 ports on each communication device. For switch A and switch B, the aforementioned port status values are obtained from these 5 ports.
[0064] The status values of the five ports of switch A are randomly concatenated with the status values of the five ports of switch B. For example, the first port status value of switch A can be concatenated with the third port status value of switch B, then the second port status value of switch A can be concatenated with the fifth port status value of switch B, and so on, until all selected port status values have been concatenated. The concatenated result serves as the edge feature between the two associated communication devices. The edge feature contains the port status information of the two communication devices and can be used for subsequent fault location and network analysis.
[0065] Figure 2 This is a schematic diagram of an undirected graph constructed based on a network, provided as an embodiment of this application. For example... Figure 2As shown, there are six communication devices R0, R1, R2, R3, R4, and R5. Communication device R0 includes ports Gi0 / 0 and Gi0 / 1, communication device R1 includes ports Gi0 / 0, Gi0 / 1, and Gi0 / 2, communication device R2 includes ports Gi0 / 0, Gi0 / 1, and Gi0 / 2, communication device R3 includes ports Gi0 / 0 and Gi0 / 1, communication device R4 includes ports Gi0 / 0, Gi0 / 1, and Gi0 / 2, and communication device R5 includes ports Gi0 / 0, Gi0 / 1, and Gi0 / 2.
[0066] Figure 3 This is a schematic diagram of undirected graph edge features provided in an embodiment of this application. For example... Figure 3 As shown, e n2 / e 21 The feature is obtained by concatenating the port states of R1-Gi0 / 1 and R2-Gi0 / 0 at both ends of the corresponding line, taking n port state values, and taking the port state of R1-Gi0 / 1. Similarly, retrieve the port status of R2-Gi0 / 0. Randomly spliced into edge features or
[0067]
[0068] 103. Based on the graph attention network algorithm, an undirected graph is modeled and the edges in the undirected graph model are classified to realize fault location of actual communication networks.
[0069] Specifically, in one embodiment of this application, the server determines multiple neighbor edge features corresponding to each edge feature based on an undirected graph; wherein, the edge connected to the corresponding node of the communication device at both ends of the edge in the undirected graph is regarded as the neighbor edge of the edge; based on the graph attention network algorithm, the edge features and the neighbor edge features corresponding to the edge features are aggregated to obtain the next layer edge features corresponding to the edge features; after the edge features are aggregated in multiple layers, the node classification representation corresponding to the multi-layer aggregated edge features is output through a fully connected layer.
[0070] In one embodiment, in the constructed undirected graph of the actual communication network, an edge feature is selected as the target edge feature. All other edge features sharing the same communication device as the target edge feature are identified; these are the neighboring edge features of the target edge feature. For the target edge feature and its corresponding neighboring edge features, an aggregation calculation is performed using a graph attention network algorithm. During the aggregation calculation, the importance coefficients between edge features are considered, the neighboring edge features are weighted and summed, and then linearly transformed and activated with the target edge feature. Through aggregation calculation, the next-layer edge features corresponding to the target edge feature are obtained.
[0071] Multiple next-layer edge features corresponding to the target edge feature are added to the target edge feature to form a new edge feature representation. This new edge feature representation is then output through a fully connected layer to obtain the final edge feature output. The above steps are repeated to aggregate and output each edge feature in the undirected graph.
[0072] In one embodiment of this application, the edge features of the next layer are calculated using the following formula:
[0073]
[0074] It should be noted that, in the embodiments of this application... Let N represent the edge feature of the i-th edge in the next layer, σ represent the activation function, K represent the number of aggregation heads in this layer, and N represent the number of aggregation heads in this layer. i W represents the characteristics of all neighboring edges of the i-th edge. k Let represent the linear transformation matrix of the edge features, and be the parameters to be learned. This represents the importance coefficient between the i-th edge and the features of each neighboring edge.
[0075] In one embodiment, in an existing undirected graph of a real communication network, all edges and their corresponding edge features are determined, and one edge is selected as the target edge. The features of all its neighboring edges are then obtained. An activation function, such as the ReLU function, is set. The number of aggregation heads in this layer, i.e., the values of the hyperparameters, is determined. The linear transformation matrix of the edge features is initialized; this is a parameter to be learned. Importance coefficients are initialized; these are also parameters to be learned, used to represent the strength of the relationship between the target edge and each neighboring edge feature.
[0076] For each neighboring edge feature of the target edge, a linear transformation matrix is used to transform the neighboring edge feature linearly. The transformed neighboring edge feature is then multiplied by the importance coefficient of the target edge to obtain a weighted neighboring edge feature. All weighted neighboring edge features are summed to obtain an aggregated neighboring edge feature. This aggregated neighboring edge feature is then concatenated with the original edge feature of the target edge, and an activation function is applied to the concatenated feature to obtain a portion of the edge feature of the next layer of the target edge.
[0077] Repeat the above steps to calculate a portion of the next-layer edge feature for each aggregation head. If there are multiple aggregation heads, their results can be concatenated or summed to obtain the final next-layer target edge feature. Add the calculated next-layer edge feature to the original edge feature of the target edge and output the updated edge feature for subsequent network fault localization tasks.
[0078] In one embodiment of this application, the server calculates the importance coefficient between an edge and the features of each neighboring edge using the following formula:
[0079]
[0080] It should be noted that in the embodiments of this application, || represents the vector concatenation operation, A k Represents the vector to be learned. This represents performing a linear transformation on the l-th layer feature vectors of i and j and then concatenating them, A k Length and The lengths are the same. A represents k and When performing the inner product, LeakyReLU() represents the activation function, and exp() represents the exponential function.
[0081] In one embodiment of this application, the server extracts features from each edge in the undirected graph model based on a pre-trained graph attention network model and classifies the extracted features; it identifies faulty edges whose classification results are faulty, and determines the two associated communication devices corresponding to the faulty edges based on the edge features of the faulty edges; it generates fault location reports corresponding to the two associated communication devices and sends the fault location reports to the corresponding management personnel to achieve fault location of the actual communication network.
[0082] In one embodiment, a pre-trained graph attention network model is used to extract features from each edge in the undirected graph model. The extracted edge features are then classified to determine the category to which each edge belongs, including normal and fault categories. Edges classified as faulty are selected from the classification results. Based on the edge features of these faulty edges, the two associated communication devices corresponding to each faulty edge are identified. Subsequently, a fault location report is generated for the two identified associated communication devices and sent to the relevant management personnel so that they can promptly understand and take measures to resolve the network fault.
[0083] Specifically, a graph attention network model is trained using historical communication network data and corresponding fault labels to accurately extract and classify edge features. Each edge in the undirected graph model is used as input, and the graph attention network model extracts features to obtain the feature representation of each edge. A classifier is used to classify the extracted edge features, determining whether each edge belongs to a fault category. Based on the classification results, edges classified as faulty are selected as faulty edges. Based on the edge features of the faulty edges, the two nodes connected to that edge in the undirected graph model are located, i.e., the associated communication devices. This generates a detailed fault location report for the associated communication devices, including fault type, fault location, and suggested solutions. The fault location report is then sent to the relevant management personnel via email, SMS, or other means.
[0084] In one embodiment of this application, the server, based on a pre-trained graph attention network model, extracts the packet count, maximum transmission unit configuration, rate configuration, MAC address, IP address, device operating temperature, system resource consumption, and on / off status of the port corresponding to each edge in the undirected graph model; inputs the edge features into the pre-trained graph attention network model to perform forward propagation calculation on the edge features, and outputs the fault probability of each edge belonging to the fault category; determines the magnitude relationship between the fault probability and a preset threshold, and determines whether each edge belongs to the fault category based on the magnitude relationship.
[0085] Locating line connection faults in a real-world communication network corresponds to classifying edges in an undirected graph. Edge category labels can be determined based on whether an edge is faulty (e.g., faulty or non-faulty), or more detailed edge classifications can be defined based on constructed edge features. For example, including packet counts as an edge feature can pinpoint lines causing congestion. Furthermore, including MTU, MAC, and IP configurations allows us to determine if the fault is caused by mismatches in these configurations.
[0086] In one embodiment, in the undirected graph model, each edge is traversed to extract the packet count, maximum transmission unit configuration, rate configuration, MAC address, IP address, device operating temperature, system resource consumption, and on / off state of the port corresponding to each edge as edge features. Then, the extracted edge features are input into a pre-trained graph attention network model. The graph attention network model performs forward propagation calculations on the edge features to obtain the fault probability of each edge belonging to a fault category.
[0087] A preset threshold is set to determine whether the failure probability of an edge is high enough to classify it as a failure category. For each edge, the failure probability corresponding to each edge is compared with the preset threshold. If the failure probability of an edge is greater than or equal to the preset threshold, the edge is determined to belong to the failure category; otherwise, the edge is determined to belong to the normal category.
[0088] Specifically, a large amount of historical communication network data and corresponding fault labels are used to train the graph attention network model, enabling it to accurately learn the relationship between edge features and fault categories, and ensuring that the features extracted from each edge are complete and accurate, so as to provide reliable information for subsequent fault judgment. The probability of each edge belonging to a fault category is calculated through the forward propagation of the graph attention network model, facilitating subsequent determination of whether an edge belongs to a fault category.
[0089] Based on actual needs and network environment, a suitable preset threshold is set to determine whether the failure probability of an edge is high enough. Then, based on the relationship between the failure probability and the preset threshold, the category of each edge is determined, and the result is output or used in subsequent fault handling processes.
[0090] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a network fault location device, the structure of which is as follows: Figure 4 As shown.
[0091] Figure 4 This is a schematic diagram of the internal structure of a network fault location device provided in an embodiment of this application. Figure 4 As shown, the device includes:
[0092] At least one processor;
[0093] And, a memory that is communicatively connected to at least one processor;
[0094] The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to:
[0095] Based on the actual communication network, identify every two associated communication devices; where associated communication devices refer to neighboring nodes that have signal transmission requirements.
[0096] The port states of two related communication devices are concatenated to obtain the corresponding edge features, and an undirected graph corresponding to the actual communication network is generated based on the edge features.
[0097] Based on the graph attention network algorithm, an undirected graph is modeled and the edges in the undirected graph model are classified to achieve fault location in actual communication networks.
[0098] This application also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, can:
[0099] Based on the actual communication network, identify every two associated communication devices; where associated communication devices refer to neighboring nodes that have signal transmission requirements.
[0100] The port states of two related communication devices are concatenated to obtain the corresponding edge features, and an undirected graph corresponding to the actual communication network is generated based on the edge features.
[0101] Based on the graph attention network algorithm, an undirected graph is modeled and the edges in the undirected graph model are classified to achieve fault location in actual communication networks.
[0102] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.
[0103] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0104] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.
[0105] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0106] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0109] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0110] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0111] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0112] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0113] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A network fault locating method, characterized by, The method includes: Based on the actual communication network, identify every two associated communication devices; wherein, the associated communication devices refer to neighboring nodes that have signal transmission requirements; The port states of two associated communication devices are concatenated to obtain the corresponding edge features, and an undirected graph corresponding to the actual communication network is generated based on the edge features. Based on the graph attention network algorithm, the undirected graph is modeled and the edges in the undirected graph model are classified to achieve fault location of the actual communication network. The step of concatenating the port states of two associated communication devices to obtain the corresponding edge features specifically includes: Determine the port status corresponding to the communication device, and obtain a preset number of port status values for each port in the two associated communication devices; wherein, the port status includes: the port's packet transmit / receive count, maximum transmission unit configuration, rate configuration, MAC address, IP address, as well as device operating temperature, system resource consumption, and on / off status; The port status values of the two associated communication devices are randomly concatenated to obtain the edge features corresponding to the edges between the two associated communication devices.
2. The method of claim 1, wherein, The step of identifying every two associated communication devices based on the actual communication network specifically includes: Obtain the actual communication network with fault location requirements, and determine all signal transmission directions in the actual communication network; Based on each signal transmission direction, the neighboring nodes with signal transmission needs are identified, and the corresponding signal senders and receivers are determined. Based on the signal sender and the signal receiver, each pair of associated communication devices in the actual communication network is identified.
3. The method of claim 1, wherein, The graph attention network-based algorithm for modeling the undirected graph specifically includes: Based on the undirected graph, multiple neighbor edge features corresponding to each edge feature are determined; wherein, the edge connected to the corresponding nodes of the communication devices at both ends of the edge in the undirected graph is regarded as the neighbor edge of the edge. Based on the graph attention network algorithm, the edge features and the neighboring edge features corresponding to the edge features are aggregated and calculated to obtain the next layer edge features corresponding to the edge features. After the edge features are aggregated in multiple layers, the node classification representations corresponding to the aggregated edge features are output through a fully connected layer.
4. The method of claim 3, wherein, The graph attention network-based algorithm aggregates the edge features and their corresponding neighboring edge features to obtain the next-layer edge features. Specifically, this includes: The edge features of the next layer are calculated using the following formula: in, Indicates the next level. Edge characteristics of each edge, This represents the activation function. The hyperparameter represents the number of aggregation heads in this layer. Indicates the first Features of all neighboring edges of an edge, Let represent the linear transformation matrix of the edge features, and be the parameters to be learned. Indicates the first The importance coefficient between each edge and the features of each neighboring edge.
5. The method of claim 4, wherein, The graph attention network-based algorithm aggregates the edge features and their corresponding neighboring edge features to obtain the next-layer edge features. Specifically, this includes: The importance coefficient between an edge and the features of each of its neighboring edges is calculated using the following formula: wherein, denotes a vector concatenation operation, denotes a vector to be learned, linearly transforms the i-th and j-th layer feature vectors and concatenates them, has the same length as has the same length as denotes and do an inner product, denotes an activation function, denotes an exponential function.
6. The network fault location method according to claim 1, characterized in that, The classification of edges in the undirected graph model to achieve fault location in the actual communication network specifically includes: Based on a pre-trained graph attention network model, features are extracted from each edge in the undirected graph model, and the extracted features are classified and judged. The fault edges that are classified as faults are identified, and based on the edge features of the fault edges, the two associated communication devices corresponding to the fault edges are determined. A fault location report is generated for the two associated communication devices, and the fault location report is sent to the corresponding management personnel to achieve fault location of the actual communication network.
7. A network fault location method according to claim 6, characterized in that, The pre-trained graph attention network model extracts features from each edge of the undirected graph model and classifies the extracted features, specifically including: Based on a pre-trained graph attention network model, in an undirected graph model, the packet count, maximum transmission unit configuration, rate configuration, MAC address, IP address, device operating temperature, system resource consumption, and on / off status of the port corresponding to each edge are extracted. The edge features are input into the pre-trained graph attention network model to perform forward propagation calculation on the edge features and output the failure probability of each edge as a failure category. The relationship between the fault probability and a preset threshold is determined, and based on this relationship, it is determined whether each edge belongs to a fault category.
8. A network fault location device, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform a network fault location method as described in any one of claims 1-7.
9. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, a network fault location method as described in any one of claims 1-7 is implemented.