A method, apparatus and system for managing network nodes

Abstract

This invention discloses a method, apparatus, and system for managing network nodes, relating to the field of computer technology. One specific embodiment of the method includes: determining a first operating value of a network node based on a first operating indicator; determining a second operating value of the network node based on the first operating indicator and a second operating indicator associated with the network node; and determining the operating status of the network node based on the first operating value and the second operating value, thereby managing the network node based on its operating status. The embodiments of this invention improve the versatility and flexibility of network node management, increase the accuracy of judging the operating status of network nodes, significantly improve the automation level of network node management, and reduce the manpower and time costs of operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus and system for managing network nodes. Background Technology

[0002] Internet application systems contain a large number of backend network nodes, and the application runs by running various types of software (such as middleware) on each network node.

[0003] Currently, the method for managing network nodes typically involves operations and maintenance personnel monitoring the values ​​of set operational indicators on the network nodes. After analyzing these values, they use pre-defined judgment strategies to determine whether there are any abnormalities in the network node's operation. However, this current method relies heavily on the experience of operations and maintenance personnel, resulting in poor flexibility and versatility. Consequently, the accuracy of judging the network node's operational status is low, leading to a waste of the human and time resources of operations and maintenance personnel. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, apparatus, and system for managing network nodes, capable of determining a first operating value of a network node based on a first operating indicator; and determining a second operating value of a network node based on the first operating indicator and a second operating indicator associated with the network node; and determining the operating status of the network node based on the first operating value and the second operating value, so as to manage the network node based on the operating status of the network node. Embodiments of the present invention improve the versatility and flexibility of network node management, improve the accuracy of judging the operating status of network nodes, significantly improve the automation level of network node management, and reduce the manpower and time costs of operation and maintenance.

[0005] To achieve the above objectives, according to one aspect of the present invention, a method for managing network nodes is provided, characterized by comprising: acquiring a first operating indicator of a network node and a second operating indicator of a plurality of other network nodes associated with the network node, wherein both the first operating indicator and the second operating indicator include multi-dimensional indicators; determining a first operating value of the network node based on the multi-dimensional indicators included in the first operating indicator; determining a second operating value of the network node based on the multi-dimensional indicators included in the first operating indicator and the multi-dimensional indicators included in the second operating indicator; and determining the operating status of the network node based on the first operating value and the second operating value, so as to manage the network node based on the operating status of the network node.

[0006] Optionally, determining the first operating value of the network node includes:

[0007] The multi-dimensional indicators included in the first operating indicator are input into a neural network model, and the neural network model is used to output the first operating value of the network node; wherein, the neural network model includes a long short-term memory network and a fully convolutional network; the first operating value is obtained by integrating the output of the long short-term memory network for the first operating indicator and the output of the fully convolutional network for the first operating indicator.

[0008] Optionally, determining the second operational value of the network node includes:

[0009] The multi-dimensional indicators included in the first operating indicators of the network node and the multi-dimensional indicators included in the second operating indicators of the plurality of other network nodes are input into the detection model, and the second operating value of the network node relative to the plurality of other nodes is output using the detection model. The detection model includes a preset dimensionality reduction model and an adaptive clustering model.

[0010] Optionally, the step of using the detection model to output the second operating value of the network node relative to the plurality of other nodes includes: using the preset dimensionality reduction model to perform dimensionality reduction operations on the multi-dimensional indicators included in the first operating indicator and the multi-dimensional indicators included in the second operating indicator, respectively, to obtain the first feature indicator of the network node and the second feature indicators corresponding to the plurality of other network nodes; inputting the first feature indicator and the plurality of second feature indicators into the adaptive clustering model to calculate the second operating value of the network node.

[0011] Optionally, the method for managing network nodes further includes: iteratively training a preset dimensionality reduction model and an adaptive clustering model using the acquired operating metrics of multiple network nodes within a historical time range; determining whether the training result of each iteration meets the iteration stopping condition for each iteration; if so, determining that the dimensionality reduction core parameters included in the training result are the dimensionality reduction core parameters possessed by the preset dimensionality reduction model; and combining the preset dimensionality reduction model with the dimensionality reduction core parameters and the adaptive clustering model to obtain the detection model.

[0012] Optionally, determining whether the training result meets the iteration stopping condition includes: comparing the evaluation values ​​of each first model included in the training result with the evaluation values ​​of the second model obtained by training the network nodes on the historical time range through the neural network model; and determining that the training result meets the iteration stopping condition if the difference between each first model evaluation value and the second model evaluation value meets the preset error range.

[0013] Optionally, the step of iteratively training the preset dimensionality reduction model and the adaptive clustering model using the acquired operating metrics of multiple network nodes within a historical time range includes: repeatedly executing the following steps N1-N3 until the iteration stopping condition is met; N1: using the preset dimensionality reduction model and the dimensionality reduction core parameters of the current cycle to perform dimensionality reduction on the operating metrics within the historical time range, obtaining the training feature metrics for the current cycle; N2: inputting the training feature metrics of the current cycle into the adaptive clustering model, and determining the first model evaluation value based on the output of the adaptive clustering model; N3: if the iteration stopping condition is not met, adjusting the dimensionality reduction core parameters of the current cycle, using the adjusted dimensionality reduction core parameters as the dimensionality reduction core parameters for the next cycle, and executing step N1.

[0014] Optionally, the adaptive clustering model includes an iterative computation model for processing features;

[0015] For each iteration of training, the method further includes: using the iterative calculation model to adaptively calculate the clustering radius of the feature data output by the preset dimensionality reduction model, outputting the clustering result based on the clustering radius, and optimizing the clustering result after the iteration stops.

[0016] Optionally, the method for managing network nodes further includes:

[0017] The following operations are performed iteratively to train the neural network model: The operating metrics of multiple network nodes within a historical time range are input into the Long Short-Term Memory (LSTM) network and the Fully Convolutional Network (WCN), respectively; the outputs of the LSM network and the WCN are combined to obtain the output of the neural network model; the training results of the neural network model are evaluated using a preset loss function and the output of the neural network model; and the neural network model is adjusted based on the evaluation results.

[0018] Optionally, the method for managing network nodes further includes:

[0019] The process involves acquiring feature data of operational metrics within the historical time range, inputting the feature data into an attention mechanism model, and using the attention mechanism model to increase the weight values ​​of feature data that are closer to the current time. The step of inputting the acquired operational metrics of multiple network nodes within the historical time range into the long short-term memory network includes: inputting the feature data processed by the attention mechanism model into the long short-term memory network to train the long short-term memory network.

[0020] Optionally, the method for managing network nodes further includes: obtaining the first operating indicator of the network node; inputting the first operating indicator into a time series prediction model, and using the time series prediction model to predict the operating status of the network node in a future time range, wherein the time series prediction model is obtained by training based on the periodic features of operating indicators in a historical time range, and periodically adjusting the prediction results of the time series prediction model.

[0021] To achieve the above objectives, according to a second aspect of the present invention, an apparatus for managing network nodes is provided, characterized in that it includes: a module for acquiring node operating indicators, a module for acquiring node operating values, and a module for determining node operating status; wherein,

[0022] The node operation indicator acquisition module is used to acquire a first operation indicator of a network node and a second operation indicator of a plurality of other network nodes associated with the network node, wherein the first operation indicator and the second operation indicator both include multi-dimensional indicators.

[0023] The node operation value acquisition module is used to determine the first operation value of the network node based on the multi-dimensional indicators included in the first operation indicator; and to determine the second operation value of the network node based on the multi-dimensional indicators included in the first operation indicator and the multi-dimensional indicators included in the second operation indicator.

[0024] The node operation status determination module is used to determine the operation status of the network node based on the first operation value and the second operation value, so as to manage the network node based on the operation status of the network node.

[0025] Optionally, the device for managing network nodes is used to determine a first operating value of the network node, comprising: inputting multi-dimensional indicators including the first operating indicator into a neural network model, and using the neural network model to output the first operating value of the network node; wherein the neural network model includes a long short-term memory network and a fully convolutional network; and integrating the output of the long short-term memory network for the first operating indicator and the output of the fully convolutional network for the first operating indicator to obtain the first operating value.

[0026] Optionally, the device for managing network nodes is used to determine the second operating value of the network node, including: inputting the multi-dimensional indicators included in the first operating indicator of the network node and the multi-dimensional indicators included in the second operating indicators of the plurality of other network nodes into a detection model, and using the detection model to output the second operating value of the network node relative to the plurality of other nodes, wherein the detection model includes a preset dimensionality reduction model and an adaptive clustering model.

[0027] Optionally, the device for managing network nodes is configured to output a second operational value of the network node relative to the plurality of other nodes using the detection model, comprising: performing dimensionality reduction operations on the multi-dimensional indicators included in the first operational indicator and the multi-dimensional indicators included in the second operational indicator using the preset dimensionality reduction model to obtain a first feature indicator of the network node and a second feature indicator corresponding to the plurality of other network nodes; inputting the first feature indicator and the plurality of second feature indicators into the adaptive clustering model to calculate the second operational value of the network node.

[0028] Optionally, the device for managing network nodes, and the method for managing network nodes, further include: iteratively training a preset dimensionality reduction model and an adaptive clustering model using acquired operating indicators of multiple network nodes within a historical time range; determining whether the training result of each iteration meets the iteration stopping condition for each iteration; if so, determining that the dimensionality reduction core parameters included in the training result are the dimensionality reduction core parameters possessed by the preset dimensionality reduction model; and combining the preset dimensionality reduction model with the dimensionality reduction core parameters and the adaptive clustering model to obtain the detection model.

[0029] Optionally, the device for managing network nodes is used to determine whether the training result meets the iteration stopping condition, including: comparing each of the first model evaluation values ​​included in the training result with the second model evaluation values ​​obtained by training the network node's operating indicators within a historical time range through a neural network model; and determining that the training result meets the iteration stopping condition if the difference between each of the first model evaluation values ​​and the second model evaluation values ​​meets a preset error range.

[0030] Optionally, the device for managing network nodes is used to iteratively train a preset dimensionality reduction model and an adaptive clustering model using the acquired operating metrics of multiple network nodes within a historical time range, including: repeatedly executing the following steps N1-N3 until the iteration stopping condition is met; N1: using the preset dimensionality reduction model and the dimensionality reduction core parameters of the current cycle to perform dimensionality reduction operation on the operating metrics within the historical time range, obtaining the training feature metrics for the current cycle; N2: inputting the training feature metrics of the current cycle into the adaptive clustering model, and determining a first model evaluation value based on the output of the adaptive clustering model; N3: if the iteration stopping condition is not met, adjusting the dimensionality reduction core parameters of the current cycle, using the adjusted dimensionality reduction core parameters as the dimensionality reduction core parameters for the next cycle, and executing step N1.

[0031] Optionally, the device for managing network nodes includes: an adaptive clustering model including an iterative computation model for processing features; for each iteration of training, it further includes: using the iterative computation model to adaptively calculate the clustering radius of the feature data output by the preset dimensionality reduction model, so as to output a clustering result based on the clustering radius, and to optimize the clustering result after the iteration stops.

[0032] Optionally, the device for managing network nodes is further configured to iteratively perform the following operations to train a neural network model: inputting the acquired operating metrics of multiple network nodes within a historical time range into the Long Short-Term Memory network and the Fully Convolutional Network respectively; combining the output of the Long Short-Term Memory network for the operating metrics within the historical time range with the output of the Fully Convolutional Network for the operating metrics within the historical time range to obtain the output of the neural network model; evaluating the training result of the neural network model using a preset loss function and the output of the neural network model; and adjusting the neural network model based on the evaluation result.

[0033] Optionally, the device for managing network nodes is further configured to acquire feature data of operating indicators within the historical time range, input the feature data into an attention mechanism model, and use the attention mechanism model to increase the weight value of feature data that is closer to the current time; the step of inputting the acquired operating indicators of multiple network nodes within the historical time range into the long short-term memory network includes: inputting the feature data processed by the attention mechanism model into the long short-term memory network to train the long short-term memory network.

[0034] Optionally, the device for managing network nodes is further configured to acquire the first operating index of the network node; input the first operating index into a time series prediction model, and use the time series prediction model to predict the operating status of the network node in the future time range, wherein the time series prediction model is trained based on the periodic features of operating indicators in the historical time range, and the prediction results of the time series prediction model are periodically adjusted.

[0035] To achieve the above objectives, according to a third aspect of the present invention, a system for managing network nodes is provided, characterized in that it includes: a plurality of network nodes, at least one of which has the means for managing network nodes as described in the second aspect; and at least any two of the plurality of network nodes are communicatively connected.

[0036] To achieve the above objectives, according to a fourth aspect of the present invention, an electronic device for managing network nodes is provided, characterized in that it includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any of the methods described above for managing network nodes.

[0037] To achieve the above objectives, according to a fifth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements any of the methods described above for managing network nodes.

[0038] One embodiment of the above invention has the following advantages or beneficial effects: it can determine a first operating value of a network node based on a first operating indicator of the network node; and determine a second operating value of the network node based on the first operating indicator and a second operating indicator of the associated network node; based on the first operating value and the second operating value, it can determine the operating status of the network node, so as to manage the network node based on the operating status of the network node. This improves the versatility and flexibility of network node management, increases the accuracy of judging the operating status of network nodes, significantly improves the automation level of network node management, and reduces the consumption of manpower and time costs in operation and maintenance.

[0039] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0040] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0041] Figure 1 This is a flowchart illustrating a method for managing network nodes according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of a process for managing network nodes according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of a training detection model provided in one embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of a device for managing network nodes according to an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of the structure of a system for managing network nodes according to an embodiment of the present invention;

[0046] Figure 6 This is an exemplary system architecture diagram in which embodiments of the present invention can be applied;

[0047] Figure 7 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0048] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0049] like Figure 1 As shown, this embodiment of the invention provides a method for managing network nodes, which may include the following steps:

[0050] Step S101: Obtain the first operating index of the network node and the second operating index of multiple other network nodes associated with the network node, wherein the first operating index and the second operating index both include multi-dimensional indicators.

[0051] Specifically, in more complex Internet application systems, the backend contains multiple network nodes, and one or more types of software run on these network nodes to jointly implement the business functions of the application system. For example, network nodes run middleware. Middleware is a type of software that sits between different parts of the application system or other application systems to achieve resource sharing and function sharing.

[0052] Furthermore, in order to determine whether there is an anomaly in the network node, the first operational indicator of the network node is obtained. The first operational indicator includes multi-dimensional indicators, such as: transaction processing volume per second, memory used, total number of keywords, traffic (inflow and / or outflow), connected client data, keyword expiration / elimination volume per second, new client connections per second, query hit rate, etc.

[0053] Furthermore, a second operating metric for multiple other network nodes associated with the aforementioned network node, similar to the first operating metric, also includes multi-dimensional metrics. Assuming an application system or multiple associated application systems have 100 network nodes: network node 0…network node 99; if network node 0 is a managed (e.g., monitored) network node, then network nodes 1 to 99 are all other network nodes associated with network node 0. This association can be direct (e.g., directly connecting to transmit data via network) or indirect (e.g., not directly connected but jointly providing configuration services). It is understood that evaluating the operating status of a network node (associated with the first operating metric) includes evaluating the network node's own operating status, as well as evaluating the network node's operating status within a group (multiple network nodes) (associated with both the first and second operating metrics).

[0054] Step S102: Determine the first operating value of the network node based on the multi-dimensional indicators included in the first operating indicator; determine the second operating value of the network node based on the multi-dimensional indicators included in the first operating indicator and the multi-dimensional indicators included in the second operating indicator.

[0055] Specifically, a first operational value for a network node is determined based on multi-dimensional indicators included in the first operational indicator obtained in real time from the network node. One embodiment for determining the first operational value is as follows: the multi-dimensional indicators included in the first operational indicator are input into a neural network model, and the neural network model outputs the first operational value of the network node. Another embodiment for determining the second operational value of the network node is as follows: the multi-dimensional indicators included in the first operational indicator of the network node, and the multi-dimensional indicators included in the second operational indicators of the plurality of other network nodes, are input into a detection model, and the detection model outputs the second operational value of the network node relative to the plurality of other nodes. The first operational value or the second operational value can be calculated based on the probability values ​​output by the model.

[0056] Furthermore, the structure and computational process of the neural network model and detection model are consistent with the descriptions in steps S201-S205, and will not be repeated here.

[0057] Step S103: Based on the first operating value and the second operating value, determine the operating status of the network node, and manage the network node based on the operating status of the network node.

[0058] Specifically, the operating status of the network node is determined based on the first operating value (e.g., obtained using a neural network model) and the second operating value (e.g., obtained using a detection model). For example, the operating status of the network node is determined using the formula a*score1+b*score2, where score1 is the first operating value and score2 is the second operating value. For example, a=0.5 and b=0.5. The calculated values ​​can be uniformly mapped to 0-100 points. It is understood that the higher the score, the more abnormal situations the network node has. This invention does not limit the specific range and value of a and b.

[0059] Further, for example, the results calculated based on the first and second running values ​​are represented by 0 to indicate that the network node is not abnormal and 1 to indicate that the network node is abnormal. For example, (0,1) indicates that the first running value is 0 and the second running value is 1; (0,0) indicates that the first running value is 0 and the second running value is also 0; and so on. Then (0,0) represents that the output of the neural network model indicates that it is normal and the output of the detection model indicates that it is normal; (0,1) represents that the output of the neural network model indicates that it is normal and the output of the detection model indicates that it is abnormal; (1,0) represents that the output of the neural network model indicates that it is abnormal and the output of the detection model indicates that it is normal; (1,1) represents that the output of both the neural network model and the detection model indicates that it is abnormal. When the output of any model is abnormal, it can be determined that the network node is abnormal. That is, it can not only detect network nodes that are not abnormal themselves but are abnormal in the group (multiple related other network nodes), but also detect network nodes that are not abnormal in the group but are abnormal themselves. For network nodes with abnormalities, it can also locate the specific indicators that cause their abnormalities, thereby improving the efficiency of operation and maintenance engineers in locating anomalies. That is, the network nodes are managed based on their operational status.

[0060] Based on the operational status of the network nodes, the network nodes are managed. For example, based on the calculated scores, they can be sorted from high to low, and the identifiers of the top set number of network nodes, as well as the specific indicator information corresponding to the anomalies, are sent to the network operation and maintenance personnel so that corresponding handling can be carried out in real time for the anomalies.

[0061] like Figure 2 As shown, this embodiment of the invention provides a process for managing network nodes, which may include the following steps:

[0062] Step S201: Obtain the multi-dimensional indicators included in the first operating indicator.

[0063] Specifically, the description of the multi-dimensional indicators included in obtaining the first operating indicator is consistent with the description of step S101, and will not be repeated here.

[0064] Step S202: Obtain multi-dimensional indicators of the second operating indicators of multiple other network nodes associated with the network node.

[0065] Specifically, the description of the multi-dimensional indicators included in obtaining the second operating indicator is consistent with the description of step S101, and will not be repeated here.

[0066] Step S203: Output the first running value using the neural network model.

[0067] Specifically, in one embodiment of the present invention, the neural network model includes a Long Short-Term Memory (LSTM) network and a Fully Convolutional Network (FCN).

[0068] Further, determining the first operating value of the network node includes: inputting the multi-dimensional indicators included in the first operating indicator into a neural network model, and using the neural network model to output the first operating value of the network node; wherein, the neural network model includes a long short-term memory network and a fully convolutional network; integrating the output of the long short-term memory network for the first operating indicator and the output of the fully convolutional network for the first operating indicator to obtain the first operating value.

[0069] Specifically, the process for training a neural network model, including a long short-term memory network and a fully convolutional network, is as follows:

[0070] 1) After obtaining multi-dimensional metrics (i.e. training data) of the operating metrics within the historical time range, the data is reconstructed. The selected metric data is processed into a set dimension (e.g., three dimensions). The sample size of the metric input each time is set to the length of the data. The number of metrics retained after feature filtering is set to the height. In order to handle the phenomenon of continuous anomalies, the average duration of anomalies is set to the width. Then the reconstructed data is input into the LSTM model and the FCN model.

[0071] 2) The training data input to the LSTM block is set to 128*10*5. Taking one sample as an example, the input data at each time step is 10*1, resulting in a 10*1 output. That is, the data structure output by the LSTM before the connection step is a 30*1 matrix. 128 is the batch size, which is the number of samples selected in one training session. After multiple training sessions, the entire dataset can be traversed. Before being input into the LSTM model, the training data is first processed by an attention mechanism model, which assigns different weights based on the distance between the time series and the current time (i.e., the time range of the detected network nodes), thus obtaining new data. In the new data, data associated with the time series receives a larger weight value. That is, the feature data processed by the attention mechanism model is input into the Long Short-Term Memory (LSTM) network to train the LSTM network. This step can increase the weight of detection indicators closer to the detection time point and decrease the weight of those farther away, thereby improving the accuracy of the neural network model in detecting the operation of network nodes.

[0072] 3) Set the training data input to the FCN block to 1*5*10, the convolution kernel to a 3*30 matrix, and the feature data to 128. Use the convolution kernel matrix to scan through the input data step by step, multiply and add the corresponding positions, and fill the input data with 0 to obtain 128 5*10 feature data.

[0073] 4) Concatenate the 10*1 data matrix output by LSTM with the 128*1 data matrix output by FCN to obtain a (10+128)*1 matrix. This integrates the outputs of the two models and outputs the predicted probability (i.e., the probability of predicting an abnormal state) through a classifier (e.g., the Softmax function). In other words, combine the output of the Long Short-Term Memory Network for the operating indicators within the historical time range with the output of the Fully Convolutional Network for the operating indicators within the historical time range to obtain the output of the neural network model.

[0074] 5) Evaluate the training results using a preset loss function, for example, the preset loss function is Loss = CrossEntropy(y',y); and use the Adam optimizer to minimize the preset loss function, while updating the training parameters of LSTM and FCN through backpropagation (i.e., adjust the neural network model according to the evaluation results; wherein, the attention mechanism model does not involve parameter updates).

[0075] Steps 1)-5) are described as follows: obtaining feature data of the operating indicators within the historical time range, inputting the feature data into the attention mechanism model, and using the attention mechanism model to increase the weight value of feature data that is closer to the current time; the step of inputting the obtained operating indicators of multiple network nodes within the historical time range into the long short-term memory network includes: inputting the feature data processed by the attention mechanism model into the long short-term memory network to train the long short-term memory network.

[0076] Further, the following operations are performed iteratively to train the neural network model: the operating metrics of multiple network nodes within a historical time range are input into the Long Short-Term Memory network and the Fully Convolutional Network, respectively; the outputs of the Long Short-Term Memory network and the Fully Convolutional Network for the operating metrics within the historical time range are combined to obtain the output of the neural network model; the training result of the neural network model is evaluated using a preset loss function and the output of the neural network model; and the neural network model is adjusted based on the evaluation result.

[0077] Step S204: Output the second running value using the detection model.

[0078] Specifically, the detection model includes a preset dimensionality reduction model and an adaptive clustering model.

[0079] In one embodiment of the present invention, the preset dimensionality reduction model is, for example, an optimized PCA (Principal Component Analysis) model, and the adaptive clustering model is, for example, an optimized DBSCAN (Density Based Spatial Clustering of Applications with Noise) model.

[0080] That is, determining the second operating value of the network node includes: inputting the multi-dimensional indicators included in the first operating indicator of the network node and the multi-dimensional indicators included in the second operating indicators of the plurality of other network nodes into a detection model, and using the detection model to output the second operating value of the network node relative to the plurality of other nodes, wherein the detection model includes a preset dimensionality reduction model and an adaptive clustering model.

[0081] Further, outputting the second operational value of the network node relative to the plurality of other nodes using the detection model includes: performing dimensionality reduction operations on the multi-dimensional indicators included in the first operational indicator and the multi-dimensional indicators included in the second operational indicator using the preset dimensionality reduction model to obtain the first feature indicator of the network node and the second feature indicators corresponding to the plurality of other nodes; inputting the first feature indicator and the plurality of second feature indicators into the adaptive clustering model to calculate the second operational value of the network node.

[0082] Preferably, the optimized preset dimensionality reduction model is a PCA model that includes a kernel function. By introducing a kernel function, most of the information of the network node's operating index data can be retained, which can better estimate the data of potential latent variables, thereby reducing the amount of computation and making the data linearly separable. It can improve the PCA model's ability to distinguish the values ​​of multi-dimensional indicators for some network nodes. Furthermore, a grid search method is used to search for the optimal kernel method and model hyperparameters. The optimal kernel method can be a polynomial kernel function selected from various kernel function methods (linear kernel, polynomial kernel function, Gaussian kernel, exponential kernel, etc.), which improves the overall model's versatility.

[0083] Furthermore, after performing dimensionality reduction on the index data using a preset dimensionality reduction model, the dimensionality-reduced feature data (including the first feature index and the second feature index) is input into the adaptive clustering model for a sequential clustering operation, thereby obtaining the second running value of the network node (i.e., the running value among multiple associated network nodes).

[0084] Preferably, two parameters can be first determined in the adaptive clustering model: epsilon: the radius of the neighboring area around a point; minPts: the minimum number of points contained in the neighboring area; according to the above two parameters and combined with the characteristics of epsilon-neighborhood, the points in the sample can be divided into three categories: core point: if NBHD(p, epsilon) >= minPts, then it is a core sample point; border point: NBHD(p, epsilon) < minPts, but this point can be obtained from some core points (density-reachable or directly-reachable); outlier: neither a core point nor a border point, then it is a point that does not belong to this category; in the case where the system is relatively complex and the number of indicators corresponding to network nodes is large, there is a problem of a large amount of computation for the existing DBSCAN model, and the value distribution of the multi-dimensional indicators of network nodes is usually uneven. For example, there is a large gap in the number of network nodes corresponding to the same application system. Therefore, the existing density-based DBSCAN algorithm has a problem of low accuracy in managing the operating conditions of network nodes; preferably, an embodiment of the present invention optimizes the existing DBSCAN algorithm to obtain an adaptive clustering model. Specifically, the optimization method is as follows: 1) Introduce the optimization method (L-BFGS), that is, an iterative calculation model. In the existing BFGS algorithm, each iterative calculation requires the Hesse matrix obtained from the previous iteration. The storage space of this Hesse matrix is at least N(N + 1) / 2, where N is the feature dimension. For higher-dimensional application scenarios, the required storage space is huge. However, the present invention uses L-BFGS to replace the previous Hesse matrix by storing a small amount of data from the previous m iterations. Therefore, using the L-BFGS optimization method (that is, the iterative calculation model for processing features) can greatly reduce the operation time and calculation cost. By this optimization method, the clustering radius of DBSCAN is changed to automatically search for the optimal radius within a small range according to the input data. That is, for the standardized data, the optimal radius is searched by the L-BFGS optimal search algorithm (that is, adaptively calculate the clustering radius of the feature data output by the preset dimensionality reduction model), so that the CH coefficient reaches the optimal (that is, the clustering result reaches the optimal); 2) Set calinski_harabasz_score as the objective function at the clustering level. Calinski_harabasz_score is a clustering index, also known as the CH score. The larger the CH score value obtained by the clustering model, the better the corresponding clustering effect. During the iterative calculation process of DBSCAN, the higher the CH score, the smaller the covariance within the category and the larger the covariance between categories, thereby improving the clustering effect and operation efficiency.

[0085] In summary, the multi-dimensional indicators included in the first operational indicator of the network node and the multi-dimensional indicators included in the second operational indicator of the plurality of other network nodes are input into the detection model (firstly, a dimensionality reduction operation is performed using a preset dimensionality reduction model, and then a clustering operation is performed using an adaptive clustering model), and the detection model is used to output the second operational value of the network node relative to the plurality of other nodes.

[0086] Step S205: Determine the operating status of the network node based on the first operating value and the second operating value.

[0087] Specifically, the description of determining the operating status of the network node based on the first operating value and the second operating value is consistent with the description of step S103, and will not be repeated here.

[0088] Step S206: Use the time series prediction model to predict the operation of the network node in the future time range.

[0089] Specifically, based on the current multi-dimensional indicator data of the network node, the time series prediction model is used to predict the operation of the network node in the future time range; specifically, the first operation indicator of the network node is obtained; the first operation indicator is input into the time series prediction model, and the time series prediction model is used to predict the operation of the network node in the future time range, wherein the prediction of the time series prediction model is obtained by training based on the periodic features of the operation indicators in the historical time range, and periodically adjusting the prediction results of the time series prediction model.

[0090] In one embodiment of the present invention, the time series prediction model can be a DeepAR model. The DeepAR model is an upgraded version of the autoregressive model, which can output a probability distribution of future data based on data within a set historical time range. Preferably, to improve the prediction effect of DeepAR, feature processing is added to the sequence data input to the model to remove noisy data and pass periodic features to the DeepAR model. That is, periodic features of the data are introduced into the training process of the DeepAR model to periodically correct the prediction results. By controlling the step size of the model fitting, the standard error of the time series prediction model is minimized, thereby improving the prediction accuracy of the time series prediction model.

[0091] like Figure 3 As shown, this embodiment of the invention provides a method for training a detection model, which may include the following steps;

[0092] Step S301: Obtain the operating metrics of multiple network nodes within a historical time range.

[0093] Step S302: Use the preset dimensionality reduction model and the core dimensionality reduction parameters of the current cycle to perform dimensionality reduction on the running indicators within the historical time range to obtain the training feature indicators for the current cycle.

[0094] Step S303: Output the second model evaluation value using the neural network model.

[0095] Step S304: Input the training feature index of the current cycle into the adaptive clustering model, and determine the first model evaluation value based on the output of the adaptive clustering model.

[0096] Step S305: Determine whether the iteration stopping condition is met. If yes, proceed to step S307; otherwise, proceed to step S306.

[0097] Step S306: Adjust the dimensionality reduction core parameters of the current cycle and use the adjusted dimensionality reduction core parameters as the dimensionality reduction core parameters of the next cycle.

[0098] Step S307: Determine the detection model, which includes the preset dimensionality reduction model and the adaptive clustering model.

[0099] Steps S301-S307 describe the process of training the detection model, namely: using the obtained operating metrics of multiple network nodes within a historical time range, the preset dimensionality reduction model and the adaptive clustering model are iteratively trained. For the training result of each iteration, it is determined whether the training result meets the iteration stopping condition. If so, the dimensionality reduction core parameters included in the training result are determined to be the dimensionality reduction core parameters possessed by the preset dimensionality reduction model. Combining the preset dimensionality reduction model with the dimensionality reduction core parameters and the adaptive clustering model, the detection model is obtained. The dimensionality reduction core parameter is, for example, K.

[0100] Further, determining whether the training result meets the iteration stopping condition includes: comparing the evaluation values ​​of each first model included in the training result with the evaluation values ​​of the second model obtained by training the network nodes on the historical time range through the neural network model; if the difference between each first model evaluation value and the second model evaluation value meets the preset error range, it is determined that the training result meets the iteration stopping condition.

[0101] Specifically, the first model evaluation value or the second model evaluation value can be the F1-score. The F1-score, for example, represents the harmonic mean of the model's precision and recall. The calculation method can be as shown in formula (1): F1 represents the F1-score.

[0102]

[0103] Where precision = TP / (TP+FP); recall = TP / (TP+FN); TP (True Positive) represents the number of positive classes predicted as positive; FN (False Negative) represents the number of positive classes predicted as negative; FP (False Positive) represents the number of negative classes predicted as positive; and TN (True Negative) represents the number of negative classes predicted as negative.

[0104] Understandably, the second model evaluation value, obtained by training the network node's operating indicators within a historical time range using a neural network model, is used as the basis for comparison with the first model evaluation value for each iteration of training the detection model. If the difference between the second model evaluation value and the first model evaluation value is within a preset error range (e.g., any value within the range of 0 to 1), and the training result is determined to meet the iteration stopping condition, then the detection model after this iteration is determined to be a well-trained detection model.

[0105] Furthermore, the iterative training of the detection model includes training a preset dimensionality reduction model and an adaptive clustering model. Specifically, historical operational index data is used as training data to iteratively train the preset dimensionality reduction model and the adaptive clustering model. The method of one iteration training is as follows: the training data is first input into the preset dimensionality reduction model to obtain the dimensionality-reduced feature index, and then the dimensionality-reduced feature index is input into the adaptive clustering model. The first model evaluation value is calculated through the output of the adaptive clustering model, and the first model evaluation value is compared with the second model evaluation value to confirm whether the iteration stopping condition is met. It can be understood that when training the detection model, the neural network model described in this invention has already been trained, that is, the second model evaluation value obtained through the neural network model can be used as the base value. The process is described by the following steps N1-N3, which involves iteratively training a preset dimensionality reduction model and an adaptive clustering model using the obtained operating metrics of multiple network nodes over a historical time range. This includes: repeatedly executing steps N1-N3 until the iteration stopping condition is met; N1: Using the preset dimensionality reduction model and the core dimensionality reduction parameters for the current cycle, perform dimensionality reduction on the operating metrics over the historical time range to obtain the training feature metrics for the current cycle; N2: Input the training feature metrics for the current cycle into the adaptive clustering model, and determine the first model evaluation value based on the output of the adaptive clustering model; N3: If the iteration stopping condition is not met, adjust the core dimensionality reduction parameters for the current cycle, use the adjusted core dimensionality reduction parameters as the core dimensionality reduction parameters for the next cycle, and execute step N1.

[0106] Furthermore, training the adaptive clustering model includes training the iterative computation model (e.g., the L-BFGS model) included in the adaptive clustering model. Specifically, training the adaptive clustering model includes: the adaptive clustering model includes an iterative computation model for processing features; for each iteration of training, it further includes: using the iterative computation model, adaptively calculating the cluster radius of the feature data output by the preset dimensionality reduction model, outputting a clustering result based on the cluster radius, and optimizing the clustering result after the iteration stops. The optimization of the adaptive clustering model is consistent with the description in step S204 and will not be repeated here; it can be understood that training the adaptive clustering model includes training all components included in the adaptive clustering model.

[0107] like Figure 4 As shown, this embodiment of the invention provides a device 400 for managing network nodes, characterized in that it includes: a node operation index acquisition module 401, a node operation value acquisition module 402, and a node operation status determination module 403; wherein,

[0108] The node operation index acquisition module 401 is used to acquire the first operation index of the network node and the second operation index of a plurality of other network nodes associated with the network node, wherein the first operation index and the second operation index both include multi-dimensional indicators.

[0109] The node operation value acquisition module 402 is used to determine the first operation value of the network node based on the multi-dimensional indicators included in the first operation indicator; and to determine the second operation value of the network node based on the multi-dimensional indicators included in the first operation indicator and the multi-dimensional indicators included in the second operation indicator.

[0110] The node operation status determination module 403 is used to determine the operation status of the network node based on the first operation value and the second operation value, so as to manage the network node based on the operation status of the network node.

[0111] like Figure 5 As shown, this embodiment of the invention provides a system 500 for managing network nodes, characterized in that it includes: a plurality of network nodes, at least one of which has the device 400 for managing network nodes; and at least any two of the plurality of network nodes are communicatively connected.

[0112] It is understandable that the device for managing network nodes can be classified as part of the network node device to be monitored, or it can be classified as part of one or more other devices.

[0113] This invention also provides an electronic device for managing network nodes, comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in any of the above embodiments.

[0114] This invention also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in any of the above embodiments.

[0115] Figure 6 An exemplary system architecture 600 is shown, which can be applied to a method or apparatus for managing network nodes according to embodiments of the present invention.

[0116] like Figure 6 As shown, system architecture 600 may include terminal devices 601, 602, and 603, a network 604, and a server 605. Network 604 serves as the medium for providing communication links between terminal devices 601, 602, and 603 and server 605. Network 604 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0117] Users can use terminal devices 601, 602, and 603 to interact with server 605 via network 604 to receive or send messages, etc. Various client applications can be installed on terminal devices 601, 602, and 603, such as e-commerce client applications, web browser applications, search applications, instant messaging tools, and email clients.

[0118] Terminal devices 601, 602, and 603 can be various electronic devices with displays and supporting various client applications, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0119] Server 605 can be a service-providing server, such as a backend management server that supports client applications used by users through terminal devices 601, 602, and 603. The backend management server can process received requests to obtain the operational status of network nodes and feed back the operational status of the network nodes corresponding to the requests to the terminal devices.

[0120] It should be noted that the method for managing network nodes provided in this embodiment of the invention is generally executed by server 605, and correspondingly, the device for managing network nodes is generally located in server 605.

[0121] It should be understood that Figure 6 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0122] The following is for reference. Figure 7 It shows a schematic diagram of the structure of a computer system 700 suitable for implementing a terminal device of the present invention. Figure 7 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0123] like Figure 7 As shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 702 or programs loaded from storage section 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the system 700. The CPU 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0124] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0125] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 709, and / or installed from removable medium 711. When the computer program is executed by central processing unit (CPU) 701, it performs the functions defined above in the system of this invention.

[0126] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0128] The modules and / or units described in the embodiments of the present invention can be implemented in software or hardware. The described modules and / or units can also be housed in a processor. For example, a processor can be described as including a module for acquiring node operating metrics, a module for acquiring node operating values, and a module for determining node operating status. The names of these modules do not necessarily limit the module itself; for example, the module for acquiring node operating metrics can also be described as "a module for acquiring a first operating metric of a network node and a second operating metric of a plurality of other network nodes associated with the network node."

[0129] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs that, when executed by the device, cause the device to: acquire a first operating indicator of a network node and second operating indicators of a plurality of other network nodes associated with the network node, wherein both the first and second operating indicators include multi-dimensional indicators; determine a first operating value of the network node based on the multi-dimensional indicators included in the first operating indicator; determine a second operating value of the network node based on the multi-dimensional indicators included in the first and second operating indicators; and determine the operating status of the network node based on the first and second operating values, so as to manage the network node based on the operating status of the network node.

[0130] Embodiments of the present invention can determine a first operating value of a network node based on a first operating indicator of the network node; and determine a second operating value of the network node based on the first operating indicator and a second operating indicator of associated network nodes; and determine the operating status of the network node based on the first operating value and the second operating value, so as to manage the network node based on the operating status of the network node. This improves the versatility and flexibility of network node management, increases the accuracy of judging the operating status of network nodes, significantly improves the automation level of network node management, and reduces the consumption of labor and time costs.

[0131] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for managing network nodes, characterized in that, include: The first operating metric of the network node and the second operating metrics of multiple other network nodes associated with the network node are obtained respectively, wherein the first operating metric and the second operating metric both include multi-dimensional metrics. Based on the multi-dimensional indicators included in the first operational indicator, a first operational value of the network node is determined; wherein, determining the first operational value of the network node includes: inputting the multi-dimensional indicators included in the first operational indicator into a neural network model, and using the neural network model to output the first operational value of the network node; wherein, the neural network model includes a long short-term memory network and a fully convolutional network; integrating the output of the long short-term memory network for the first operational indicator and the output of the fully convolutional network for the first operational indicator to obtain the first operational value. Based on the multi-dimensional indicators included in the first operating indicator and the multi-dimensional indicators included in the second operating indicator, a second operating value of the network node is determined; wherein, determining the second operating value of the network node includes: inputting the multi-dimensional indicators included in the first operating indicator of the network node and the multi-dimensional indicators included in the second operating indicators of the plurality of other network nodes into a detection model, and using the detection model to output the second operating value of the network node relative to the plurality of other network nodes; Based on the first operating value and the second operating value, the operating status of the network node is determined, and the network node is managed based on the operating status of the network node.

2. The method for managing network nodes according to claim 1, characterized in that, The detection model includes a preset dimensionality reduction model and an adaptive clustering model.

3. The method according to claim 2, characterized in that, The step of using the detection model to output a second operational value of the network node relative to the plurality of other network nodes includes: The preset dimensionality reduction model is used to perform dimensionality reduction operations on the multi-dimensional indicators included in the first operating indicator and the multi-dimensional indicators included in the second operating indicator, respectively, to obtain the first feature indicator of the network node and the second feature indicators corresponding to multiple other network nodes. The first feature index and multiple second feature indices are input into the adaptive clustering model to calculate the second operating value of the network node.

4. The method according to claim 3, characterized in that, Also includes: The preset dimensionality reduction model and adaptive clustering model are iteratively trained using the operating indicators of multiple network nodes within a historical time range. For the training results of each iteration, it is determined whether the training results meet the iteration stopping condition. If so, the dimensionality reduction core parameters included in the training results are determined to be the dimensionality reduction core parameters of the preset dimensionality reduction model. The detection model is obtained by combining the preset dimensionality reduction model with the core dimensionality reduction parameters and the adaptive clustering model.

5. The method according to claim 4, characterized in that, The step of determining whether the training result meets the iteration stopping condition includes: The training results include the evaluation values ​​of each first model, which are compared with the evaluation values ​​of the second model obtained by training the network nodes with the operating indicators of the network nodes in the historical time range through the neural network model. If the difference between the evaluation values ​​of the first model and the evaluation values ​​of the second model meets the preset error range, the training result is determined to meet the iteration stopping condition.

6. The method according to claim 4, characterized in that, The step of iteratively training the preset dimensionality reduction model and adaptive clustering model using the acquired operating metrics of multiple network nodes over a historical time range includes: Repeat steps N1-N3 until the iteration stopping condition is met; N1: Using the preset dimensionality reduction model and the core dimensionality reduction parameters of the current cycle, perform dimensionality reduction operation on the running indicators within the historical time range to obtain the training feature indicators for the current cycle. N2: Input the training feature index of the current cycle into the adaptive clustering model, and determine the first model evaluation value based on the output of the adaptive clustering model; N3: If the iteration stopping condition is not met, adjust the dimensionality reduction core parameters of the current cycle, use the adjusted dimensionality reduction core parameters as the dimensionality reduction core parameters of the next cycle, and execute step N1.

7. The method according to claim 4, characterized in that, The adaptive clustering model includes an iterative computation model for processing features; For each training iteration, it also includes: Using the iterative calculation model, the clustering radius of the feature data output by the preset dimensionality reduction model is adaptively calculated, and the clustering result is output based on the clustering radius. After the iteration stops, the clustering result is optimized.

8. The method according to claim 1, characterized in that, Also includes: The following operations are performed iteratively to train the neural network model: The operating metrics of multiple network nodes within the historical time range are respectively input into the Long Short-Term Memory Network and the Fully Convolutional Network. The output of the Long Short-Term Memory Network for the operating metrics within the historical time range is combined with the output of the Fully Convolutional Network for the operating metrics within the historical time range to obtain the output of the neural network model. The training results of the neural network model are evaluated using a preset loss function and the output of the neural network model. The neural network model is adjusted based on the evaluation results.

9. The method according to claim 8, characterized in that, Further including: The feature data of the operating indicators within the historical time range are obtained, and the feature data are input into the attention mechanism model. The attention mechanism model is used to increase the weight value of the feature data that is closer to the current time. The step of inputting the acquired operational metrics of multiple network nodes within a historical time range into the Long Short-Term Memory network includes: The feature data processed by the attention mechanism model is input into the long short-term memory network to train the long short-term memory network.

10. The method according to claim 1, characterized in that, Further including: Obtain the first operational metric of the network node; The first operational indicator is input into a time series prediction model, which is then used to predict the operational status of the network node within a future time range. The time series prediction model is trained based on the periodic features of operational indicators within a historical time range, and the prediction results are periodically adjusted.

11. An apparatus for managing network nodes, characterized in that, include: The system includes a module for obtaining node performance metrics, a module for obtaining node performance values, and a module for determining node performance status. The node operation indicator acquisition module is used to acquire a first operation indicator of a network node and a second operation indicator of a plurality of other network nodes associated with the network node, wherein the first operation indicator and the second operation indicator both include multi-dimensional indicators. The node operation value acquisition module is used to determine a first operation value of the network node based on the multi-dimensional indicators included in the first operation indicator; wherein, determining the first operation value of the network node includes: inputting the multi-dimensional indicators included in the first operation indicator into a neural network model, and using the neural network model to output the first operation value of the network node; wherein, the neural network model includes a long short-term memory network and a fully convolutional network; integrating the output of the long short-term memory network for the first operation indicator and the output of the fully convolutional network for the first operation indicator to obtain the first operation value; determining a second operation value of the network node based on the multi-dimensional indicators included in the first operation indicator and the multi-dimensional indicators included in the second operation indicator; wherein, determining the second operation value of the network node includes: inputting the multi-dimensional indicators included in the first operation indicator of the network node and the multi-dimensional indicators included in the second operation indicators of the plurality of other network nodes into a detection model, and using the detection model to output the second operation value of the network node relative to the plurality of other network nodes; The node operation status determination module is used to determine the operation status of the network node based on the first operation value and the second operation value, so as to manage the network node based on the operation status of the network node.

12. A system for managing network nodes, characterized in that, include: Multiple network nodes, at least one of which has the means for managing network nodes as described in claim 11; At least two of the plurality of network nodes are in communication connection.

13. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-10.

14. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-10.

Images