Microservice system root cause positioning method and system based on spatio-temporal graph neural network

Abstract

The application belongs to the technical field of big data information processing, and provides a micro-service system root cause positioning method and system based on a space-time graph neural network, acquires calling data information and monitoring index data of a micro-service system, constructs a micro-service time sequence calling topology graph stack, combines the topology graph and monitoring index information, uses a long short-term memory artificial neural network model to learn time sequence feature information of the topology graph, and uses a graph convolutional neural network model to learn spatial structure feature information of the topology graph; the learned time sequence feature information and spatial structure feature information are subjected to space-time feature fusion, and an attention mechanism is introduced during the space-time feature fusion; based on the space-time fusion feature of each topology graph node after fusion, a root cause positioning model is used to determine the probability of each node being a root cause, and a final result of root cause positioning is obtained. The application can fully mine space-time feature information in a topology graph, and effectively perform root cause positioning, thereby improving the accuracy of root cause positioning.

Description

Technical Field

[0001] This invention belongs to the field of big data information processing technology, specifically relating to a root cause localization method and system for microservice systems based on spatiotemporal graph neural networks. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the development of software development technologies, microservice architecture, as a new design concept, has been widely applied to online service systems. Microservice architecture decomposes online service systems into smaller and more reusable components, offering high flexibility and accelerating software development. In traditional monolithic architectures, applications can run in a single process, easily maintaining application availability. However, microservice architecture, due to its unique characteristics, requires collaboration between reusable components. When a single service fails, a cascading effect can occur, impacting the entire application. Therefore, fault localization is crucial in microservice systems.

[0004] With the continuous development of artificial intelligence (AI) technology, more and more research is combining AI with root cause localization methods to more accurately pinpoint root causes. In recent years, topology-based root cause localization methods have been widely proposed. This method mainly constructs a topology graph based on the call and attribution relationships between microservices in a system, and then explores the root cause by analyzing the topology graph to obtain the root cause service. However, the metrics data of microservices and between microservices exhibit temporal sequence characteristics. Therefore, how to simultaneously learn spatial structural features and temporal sequence characteristics within the topology using AI methods has become a key research focus. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a root cause localization method and system for microservice systems based on spatiotemporal graph neural networks. This invention can fully mine the spatiotemporal feature information in the topology graph and perform effective root cause localization, thereby improving the accuracy of root cause localization.

[0006] According to some embodiments, the present invention adopts the following technical solution:

[0007] A root cause localization method for microservice systems based on spatiotemporal graph neural networks includes the following steps:

[0008] Obtain call data and monitoring metrics data from the microservice system;

[0009] The acquired data is preprocessed, and a microservice time-series call topology stack is constructed based on the preprocessed data. The microservice time-series call topology stack includes multiple topology graphs.

[0010] By combining topology graphs and monitoring indicator information, a long short-term memory artificial neural network model is used to learn the time series features of the topology graph, and a graph convolutional neural network model is used to learn the spatial structure features of the topology graph.

[0011] The learned time series features and spatial structure features are fused into spatiotemporal features, and an attention mechanism is introduced during the spatiotemporal feature fusion to calculate the attention weights of the time series and spatial structure features.

[0012] Based on the spatiotemporal fusion characteristics of each node in the fused topology graph, the probability of each node being a root cause is determined using the root cause localization model, thus obtaining the final result of root cause localization.

[0013] As an alternative implementation, the call data information includes the call relationship between microservices, the monitoring indicator data includes the monitoring indicators of microservices and the monitoring indicator information between microservices, the monitoring indicators include several of CPU utilization, memory utilization, response time, number of requests and error rate, and the monitoring indicator information includes several of call latency, call frequency and failure rate.

[0014] As an alternative implementation method, the process of preprocessing the acquired data includes: data cleaning and missing data completion, as well as normalization of the monitoring indicator data to obtain feature representation vectors.

[0015] As an alternative implementation, the process of constructing a microservice time-series call topology graph stack based on preprocessed data includes: dividing the abnormal time period into multiple time intervals; constructing a service call topology graph based on the call information between microservices within each time interval; and constructing a service call topology graph consisting of a set of microservice nodes V = {v1, v2, ..., v...}. s Composed of}, during an abnormal time period, when microservice v i Towards microservices j When sending a call request, add a microservice v i Point to v j The directed edges are used to construct a service call topology graph G = (V, E, A) at each time interval, where V represents a node in the graph, which is a set of microservices, E represents an edge, and A represents an adjacency matrix. During abnormal time periods, the n constructed service call topology graphs constitute a microservice time-series call topology stack. n is the number of time intervals.

[0016] Furthermore, in the microservice sequential call topology stack, This is used to represent the feature matrix of microservices within a time interval, where S represents the number of nodes and C represents the number of microservice monitoring metrics.

[0017] Will The adjacency matrix used to represent time intervals is represented in the service call topology graph as follows:

[0018]

[0019] Based on inter-microservice call latency call frequency and failure rate Calculate the weight matrix W of the edges t microservices v i and v j Weight matrix between Represented as:

[0020]

[0021] Where α, β, and γ are weighting parameters.

[0022] As an alternative implementation, during the process of learning the spatial structure features of the topological graph using a graph convolutional neural network model, the weight W of the edges between microservices is used. t Update adjacency matrix A t ,get

[0023] Update the adjacency matrix and microservice node monitoring metrics characteristics X t As input, a graph convolutional neural network model is used to learn spatial structure feature representations.

[0024] Where σ is the sigmoid function, W GCN These are the learnable weights of a graph convolutional neural network model.

[0025] As an alternative implementation method, the process of spatiotemporal feature fusion of the learned time series feature information and spatial structure feature information includes: performing feature alignment operation on the time series features and spatial structure features using linear transformation;

[0026] Based on the aligned time series feature matrix, query, key and value vectors are generated respectively, and attention weights of time series features are calculated using an attention mechanism.

[0027] Based on the aligned spatial structure feature matrix, query, key, and value vectors are generated respectively, and attention weights of spatial structure features are calculated using an attention mechanism.

[0028] The spatiotemporal feature weights are calculated based on the obtained spatiotemporal feature attention weights, and multiple attention heads are used to fuse spatiotemporal features.

[0029] As an alternative implementation, the root cause localization model is a binary classification model, based on the spatiotemporal fusion characteristics of each topology graph node. Output node at t n The probability of being the root cause node within a time interval.

[0030]

[0031] in, Both b are learnable parameters, and σ is the sigmoid function;

[0032] Node v i In all time intervals t = {t1, t2, ..., t n The probabilities in} are summed to obtain the final root cause probability:

[0033] Rank the root cause probabilities of all nodes to obtain the final root cause microservice.

[0034] A root cause localization system for microservice systems based on spatiotemporal graph neural networks, comprising:

[0035] The data acquisition module is used to acquire call data information and monitoring metric data of the microservice system;

[0036] The timing call topology graph stack construction module is used to preprocess the acquired data and construct a microservice timing call topology graph stack based on the preprocessed data. The microservice timing call topology graph stack includes multiple topology graphs.

[0037] The spatiotemporal feature learning module is used to combine topology graph and monitoring indicator information, using a long short-term memory artificial neural network model to learn the time series features of the topology graph, and using a graph convolutional neural network model to learn the spatial structure features of the topology graph.

[0038] The spatiotemporal feature fusion module is used to perform spatiotemporal feature fusion on the learned time series feature information and spatial structure feature information, and introduces an attention mechanism during spatiotemporal feature fusion to calculate the attention weights of time series and spatial structure features;

[0039] The root cause localization module is used to determine the probability of each node being a root cause based on the spatiotemporal fusion features of each node in the fused topology graph, and obtain the final result of root cause localization.

[0040] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the steps in the above method.

[0041] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps in the method described above.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] (1) This invention fully considers the temporal sequence between monitoring indicators, collects and processes the call information and monitoring indicator data of the microservice system, sets a specific time interval, and constructs the microservice system temporal call topology stack.

[0044] (2) This invention fully learns the time series and spatial structure features of the topological graph through long short-term neural network model and graph neural network, and introduces attention mechanism to perform spatiotemporal feature fusion to improve the accuracy of root cause localization.

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0046] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0047] Figure 1 This is an overall flowchart of a microservice system root cause localization method based on a spatiotemporal graph neural network in one embodiment;

[0048] Figure 2 This is a data flow processing diagram in one embodiment;

[0049] Figure 3 This is a flowchart illustrating the method execution in one embodiment;

[0050] Figure 4 This is a schematic diagram of the root cause localization system of a microservice system based on a spatiotemporal graph neural network in one embodiment. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] Where there is no conflict, the embodiments and features described in this application may be combined with each other.

[0055] Example 1

[0056] like Figure 3 As shown, this embodiment provides a root cause localization method for microservice systems based on spatiotemporal graph neural networks, including the following steps:

[0057] Step 1: Obtain call information and monitoring metrics data from the microservice system;

[0058] In this embodiment, the collection of call information and monitoring metrics data of the microservice system is carried out using Prometheus to collect call information and monitoring metrics information between microservices during system runtime, and data preprocessing is performed, including data cleaning, missing data completion, data definition and storage.

[0059] Of course, in other embodiments, the monitoring tools are not limited to the examples given above, such as... Figure 1 As shown, the call information and monitoring metrics of the microservice system can be obtained directly from the relevant data sources.

[0060] Similarly, in some embodiments, call information and monitoring indicator data can also be obtained from other sources.

[0061] In this embodiment, the call information of the microservice system refers to the call relationships between microservices, and the monitoring metrics are divided into microservice monitoring metrics and inter-microservice monitoring metrics. Microservice monitoring metrics include, but are not limited to, CPU utilization, memory utilization, response time, number of requests, and error rate, while inter-microservice monitoring metrics include, but are not limited to, call latency, call frequency, and failure rate.

[0062] Step 2: Perform preprocessing operations on the collected microservice system monitoring data to obtain feature representation vectors;

[0063] Step 2 involves preprocessing the microservice system monitoring metrics data, specifically including:

[0064] Because monitoring metrics in microservice systems often have different dimensions and units, this can impact subsequent root cause analysis. Therefore, to standardize the dimensions of monitoring metrics across microservice systems, a min-max normalization method is used to normalize the monitoring metric data.

[0065]

[0066] Here, x represents a monitoring metric in the microservice system, min() takes the minimum value of metric x, max() takes the maximum value of metric x, and ε is a very small constant to avoid the denominator being 0. The final value range of the monitoring metric is [0,1].

[0067] Step 3: Based on the call information between microservices, construct the microservice time-series call topology stack according to the time interval;

[0068] Step 301: In this embodiment, the abnormal time period t is divided into n time intervals, t = {t1, t2, ..., tn}. n}

[0069] Within each time interval, a service call topology graph is constructed based on the call information between microservices. The service call topology graph consists of a set of microservice nodes V = {v1, v2, ..., v...} s Composed of}, within the time period t, when microservice v i Towards microservices j When sending a call request, add a microservice v i Point to v j The directed edges are then used. Finally, a service call topology graph G = (V, E, A) is constructed at each time interval, where V represents the nodes of the graph, a set of microservices, E represents the edges, and A represents the adjacency matrix. During abnormal time periods, the constructed n service call topology graphs constitute the microservice sequential call topology stack.

[0070] Step 302: In the microservice sequential call topology stack, This is used to represent the feature matrix of a microservice within a time interval, where S represents the number of nodes and C represents the number of microservice monitoring metrics.

[0071] Will The adjacency matrix used to represent time intervals can be represented in the service call topology graph as follows:

[0072]

[0073] Step 303: The correlation strength of monitoring metric data between microservices determines the correlation weight; the higher the weight, the stronger the correlation. Therefore, based on the call latency between microservices... call frequency and failure rate Calculate the weight matrix W of the edges t Microservices v i and v j The weight matrix between them can be represented as:

[0074]

[0075] Where α, β, and γ are weighting parameters.

[0076] Step 4: Learn the temporal feature information of each topology graph based on the microservice temporal call topology stack, such as... Figure 2 As shown;

[0077] In this embodiment, the topology graph G in the microservice time-series call topology stack is defined by time interval t. t For example, using a Long Short-Term Memory (LSTM) network to learn G... t The specific steps for obtaining time series feature information are as follows:

[0078] i t =σ(W i ·[h t-1 ,X t ]+b i )

[0079] f t =σ(W f ·[h t-1 ,X t ]+b f )

[0080]

[0081]

[0082] o t =σ(W o ·[h t-1 ,X t ]+b o )

[0083] h t =o t ⊙tanh(c t )

[0084] Among them, h t It is in a hidden state, c t In cellular state, f t i t o tThese represent the forget gate, input gate, and output gate, respectively. Denotes candidate units, σ ​​is the sigmoid function, ⊙ denotes element-wise multiplication, and W f W i W c W o Both represent learnable weights, b i b f b c b o This represents the paranoia vector.

[0085] The above steps can be simplified as follows:

[0086] H t =LSTM(X) t )

[0087] Finally, the time series feature representation of the topological graph is obtained. d T The feature dimension of the LSTM output represents the dimension of the time feature.

[0088] Step 5: Learn the spatial structure feature information of each topology graph based on the microservice time-series call topology stack;

[0089] In step 5, the topology graph G in the microservice time-series call topology stack is the result of time interval t. t For example, learning spatial structure feature representations in topological graphs using Graph Convolutional Networks (GCNs) includes:

[0090] Step 501: In order to fully capture the spatial structure information of the topology graph, based on the weight W of the edges between microservices... t Update adjacency matrix A t ,get

[0091]

[0092] Step 502: Update the adjacency matrix and microservice node monitoring metrics characteristics X t As input, a graph convolutional network (GCN) is used to learn spatial structure feature representations.

[0093]

[0094] Where σ is the sigmoid function, W GCN Denotes the learnable weights, d S Dimensions that represent spatial features.

[0095] Step 6: Based on spatiotemporal feature information, perform spatiotemporal feature fusion using an attention mechanism to obtain a spatiotemporal fusion feature vector;

[0096] In step 6, the learned time series features and spatial structure features Feature fusion includes:

[0097] Step 601: Since time series features and spatial structure features may have different feature dimensions, a linear transformation is used to align the spatiotemporal features:

[0098]

[0099]

[0100] in, Both are parameter matrices, and the final result is the feature-aligned matrix. and d F This represents the aligned feature dimensions.

[0101] Step 602: Based on the aligned time series feature matrix, generate query Q respectively. t Key K t Value vector V t The attention weight α of time series features is calculated using an attention mechanism. T :

[0102]

[0103]

[0104]

[0105]

[0106] in, Denotes the weight parameter matrix, d K This represents the attention dimension.

[0107] Step 603: Based on the aligned spatial structure feature matrix, generate query, key, and value vectors respectively, and use an attention mechanism to calculate the spatial structure feature attention weight α. S :

[0108]

[0109]

[0110]

[0111]

[0112] in, This represents the weight parameter matrix.

[0113] Step 604: Calculate the spatiotemporal feature weights based on the obtained spatiotemporal feature attention weights, and perform spatiotemporal feature fusion using h attention heads:

[0114] T at =α T ·V t

[0115] S at =α S ·V s

[0116]

[0117] Among them, T at and S at W represents the weighted value of time series features and the weighted value of spatial structure features, respectively. at Let F be the projection matrix. at The spatiotemporal characteristics after final fusion.

[0118] Step 7: Train the root cause localization model to obtain the root cause localization results of the microservice system.

[0119] In this embodiment, root cause localization is designed as a binary classification model, with the spatiotemporal fusion features of each topology graph node obtained in step 6 as input. Output node at t n The probability of being the root cause node within a time interval.

[0120]

[0121] in, Both b and are learnable parameters.

[0122] Since a time-series call topology stack is constructed in step 3, node v i In all time intervals t = {t1, t2, ..., t n The probabilities in} are summed to obtain the final root cause probability:

[0123]

[0124] Finally, the root cause probabilities of all nodes are ranked to obtain the final root cause microservice.

[0125] Example 2

[0126] like Figure 4 As shown, this embodiment provides a microservice system root cause localization system based on spatiotemporal graph neural networks, including:

[0127] The data acquisition module is used to acquire call data information and monitoring metric data of the microservice system;

[0128] The timing call topology graph stack building module is used to build a microservice timing call topology graph stack based on the call data information and monitoring metric data of the microservice system.

[0129] The time-series feature learning module is used to learn the time-series feature information of monitoring indicators in the topology graph based on LSTM.

[0130] The spatial feature learning module is used to learn the spatial structural feature information of monitoring indicators in the topology map based on GCN.

[0131] The spatiotemporal feature fusion module is used for spatiotemporal feature fusion based on the attention mechanism, and learns its fused features for subsequent root cause localization.

[0132] The root cause localization module is used to obtain root cause localization results using the root cause localization model.

[0133] Example 3

[0134] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the root cause localization method for a microservice system based on a spatiotemporal graph neural network provided in Embodiment 1.

[0135] Example 4

[0136] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the root cause localization method for a microservice system based on a spatiotemporal graph neural network provided in Embodiment 1 above.

[0137] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention 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.

[0138] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. 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 illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0139] 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.

[0140] 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.

[0141] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made by those skilled in the art without creative effort within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A root cause localization method for microservice systems based on spatiotemporal graph neural networks, characterized in that, Includes the following steps: Obtain call data and monitoring metrics data of the microservice system; the call data includes the call relationship between microservices, and the monitoring metrics data includes the monitoring metrics of microservices and the monitoring metrics information between microservices. The monitoring metrics include several of the following: CPU utilization, memory utilization, response time, number of requests, and error rate. The monitoring metrics information includes several of the following: call latency, call frequency, and failure rate. The acquired data is preprocessed, and based on the preprocessed data, a microservice time-series call topology stack is constructed. This includes dividing the abnormal time period into multiple time intervals, and within each time interval, constructing a service call topology graph based on the call information between microservices. The service call topology graph consists of a set of microservice nodes. Composition, during abnormal time periods, when microservices Towards microservices Add a microservice when sending the call request. point to The directed edges are used to construct a service call topology graph in each time interval. Where V represents a node in the graph, which is a set of microservices, E represents an edge, and A represents an adjacency matrix; during abnormal time periods, the constructed n service call topology graphs constitute the microservice time-series call topology stack. n is the number of time intervals, and the microservice time-series call topology stack includes multiple topology graphs; By combining topology graph and monitoring indicator information, a long short-term memory artificial neural network model is used to learn the time series features of the topology graph, and a graph convolutional neural network model is used to learn the spatial structure features of the topology graph. The process of spatiotemporal feature fusion of learned time series features and spatial structure features includes: performing feature alignment operations on time series features and spatial structure features using linear transformations. ;in, , Both are parameter matrices, and the final result is the feature-aligned matrix. and , Indicates the aligned feature dimensions; An attention mechanism is introduced during spatiotemporal feature fusion to calculate the attention weights for time series and spatial structure features: based on the aligned time series feature matrix, queries are generated respectively. ,key Value vector The attention mechanism is used to calculate the attention weights of time series features. : ; ; ; ;in, , , Represents the weight parameter matrix, Indicates the attention dimension; Based on the aligned spatial structure feature matrix, query, key, and value vectors are generated respectively, and attention weights for spatial structure features are calculated using an attention mechanism. : ; ; ; ;in, , , Represents the weight parameter matrix; Based on the obtained spatiotemporal feature attention weights, a spatiotemporal feature weighted value is calculated, and multiple attention heads are used to fuse spatiotemporal features: ; ; ;in, and These are the weighted values ​​of time series features and the weighted values ​​of spatial structure features, respectively. Let be the projection matrix. The spatiotemporal characteristics after final fusion; Based on the spatiotemporal fusion characteristics of each node in the fused topology graph, the probability of each node being a root cause is determined using the root cause localization model, thus obtaining the final result of root cause localization.

2. The root cause localization method for microservice systems based on spatiotemporal graph neural networks as described in claim 1, characterized in that, The preprocessing of the acquired data includes: data cleaning and missing data completion, as well as normalization of the monitoring indicator data to obtain feature representation vectors.

3. The root cause localization method for microservice systems based on spatiotemporal graph neural networks as described in claim 1, characterized in that, In the microservice sequential call topology stack, This is used to represent the feature matrix of microservices within a time interval, where S represents the number of nodes and C represents the number of microservice monitoring metrics. Will The adjacency matrix used to represent time intervals is represented in the service call topology graph as follows: ； Based on inter-microservice call latency Call frequency and failure rate Calculate the edge weight matrix microservices and Weight matrix between Represented as: ； in, , and These are the weight parameters.

4. The root cause localization method for microservice systems based on spatiotemporal graph neural networks as described in claim 1, characterized in that, In the process of learning the spatial structure features of a topological graph using a graph convolutional neural network model, the weights of edges between microservices are used. Update the adjacency matrix ,get : ; Update the adjacency matrix and microservice node monitoring metrics characteristics As input, a graph convolutional neural network model is used to learn spatial structure feature representations. : ； in, For the sigmoid function, For the learnable weights of a graph convolutional neural network model, Dimensions that represent spatial features.

5. The root cause localization method for microservice systems based on spatiotemporal graph neural networks as described in claim 1, characterized in that, The root cause localization model is a binary classification model, based on the spatiotemporal fusion characteristics of each topology graph node. The output node is The probability of being the root cause node within a time interval. : ； in, Both b and are learnable parameters. It is the sigmoid function; Node In all time intervals The probabilities in each case are summed to obtain the final root cause probability: ; Rank the root cause probabilities of all nodes to obtain the final root cause microservice.

6. A microservice system root cause localization system based on spatiotemporal graph neural network, employing the microservice system root cause localization method based on spatiotemporal graph neural network as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire call data information and monitoring metric data of the microservice system; The timing call topology graph stack construction module is used to preprocess the acquired data and construct a microservice timing call topology graph stack based on the preprocessed data. The microservice timing call topology graph stack includes multiple topology graphs. The temporal feature learning module is used to learn the time-series features of the topology map by combining the topology map and monitoring indicator information and using a long short-term memory artificial neural network model. The spatial feature learning module is used to learn the spatial structural features of the topology map by combining the topology map and monitoring indicator information and using a graph convolutional neural network model. The spatiotemporal feature fusion module is used to perform spatiotemporal feature fusion on the learned time series feature information and spatial structure feature information, and introduces an attention mechanism during spatiotemporal feature fusion to calculate the attention weights of time series and spatial structure features; The root cause localization module is used to determine the probability of each node being a root cause based on the spatiotemporal fusion features of each node in the fused topology graph, and obtain the final result of root cause localization.

7. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the processor executes the computer instructions, they complete the steps in the root cause localization method for a microservice system based on a spatiotemporal graph neural network as described in any one of claims 1-5.

Images