An observable analysis method and system based on intelligent analysis

Abstract

The present invention discloses an observable analysis method and system based on intelligent analysis, the method comprising the following steps: a collector module collects operation and maintenance data and sends it to an observable module, the operation and maintenance data including client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data; the observable module performs data analysis based on the operation and maintenance data, the data analysis including root cause location, performance optimization, capacity planning and prediction, operation and maintenance release process monitoring, microservice governance and user experience optimization; the observable module sends the results of the data analysis to an output module. The observable analysis method proposed in the present invention ensures full-link monitoring from the end user to the back-end server, forms a global system view, and can quickly locate the root cause of the fault, reducing the time for manual troubleshooting.

Description

Technical Field

[0001] The present invention belongs to the technical field of system operation and maintenance, and in particular relates to an observable analysis method and system based on intelligent analysis. Background Art

[0002] Observability, as a modern operation and maintenance concept, emphasizes that the system should have comprehensive, in-depth and understandable state acquisition capabilities during operation. By collecting and analyzing various observable data of the system and building a comprehensive and three-dimensional monitoring and analysis system, the operation and maintenance team can understand the health status, performance and fault causes of the system in real time in a complex and dynamic IT environment, and make accurate decisions based on this information to achieve rapid problem location, preventive maintenance and continuous optimization.

[0003] The Chinese invention patent application with publication number CN118733395A discloses a method and system for integrated visualization monitoring of observable data of a distributed system, which realizes data model unification through standard semantic conventions and standard data transmission protocols, and tracks the flow path of requests between various services and records telemetry information by assigning a unique identifier to each request and passing the identifier in the system, and performs real-time monitoring and historical data query through a monitoring dashboard; the system includes an S201 data collection module, an S202 standardized processing module, an S203 data transmission module, an S204 data processing center and an S205 monitoring dashboard, which can automatically collect, integrate and analyze key data in a distributed system to achieve comprehensive monitoring and in-depth insight into the system status.

[0004] Existing traditional monitoring tools mainly rely on single-level data collection and cannot provide cross-level, in-depth system observability analysis, which makes it impossible for operation and maintenance personnel to quickly and accurately identify the root cause of the problem, optimize performance, and improve user experience. It also faces the following challenges: insufficient intelligent analysis capabilities. Traditional monitoring tools can only provide basic indicator monitoring and log analysis, lacking intelligent analysis capabilities, resulting in low efficiency in root cause location and performance optimization; insufficient multi-dimensional data integration. Existing monitoring methods cannot effectively integrate data from the application layer, network layer, and infrastructure layer, making it difficult to form a global system view and unable to support cross-level analysis and decision-making; inaccurate capacity forecasting. Due to the lack of integrated analysis of multi-dimensional data, the system's capacity forecasting is often not accurate enough, resulting in insufficient or excessive resource allocation. Summary of the invention

[0005] The present invention provides an observable analysis method and system based on intelligent analysis, integrating multi-dimensional data from the application layer, network layer and infrastructure layer, realizing multiple observable capabilities, focusing on intelligent analysis capabilities, ensuring the ability to identify faults in real time, optimize system performance and accurately predict capacity requirements, aiming to solve the problems existing in the prior art of being unable to provide cross-level and in-depth system observability analysis and insufficient intelligent analysis capabilities.

[0006] In order to solve the above technical problems, the observable analysis method proposed in the present invention comprises the following steps:

[0007] The collector module collects operation and maintenance data and sends it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data;

[0008] The observability module performs data analysis based on operation and maintenance data, including root cause location, performance optimization, capacity planning and forecasting, operation and maintenance release process monitoring, microservice governance, and user experience optimization;

[0009] The observability module sends the results of data analysis to the output module.

[0010] Preferably, the collector module includes:

[0011] The log collector further includes a client log collector that monitors the client operation behavior and a server log collector that collects the server application log;

[0012] Cluster collector, used to collect cluster performance data and container performance data;

[0013] Infrastructure collector, used to collect server performance data, database performance data and gateway performance data.

[0014] Preferably, the data collection method of the collector module includes active collection and passive reporting, wherein the active collection obtains real-time data through the Prometheus platform deployed in the server, and the passive reporting obtains log data through the Clickhouse database;

[0015] The collector module also performs data conversion on the passively reported data, and the conversion method is:

[0016] Extract timestamps and indicator data from the Clickhouse database and convert them into time series data specified by Prometheus;

[0017] Map data fields, map timestamp fields to Prometheus time series timestamps, convert different dimensions or fields to Prometheus labels, and convert numeric fields to the values ​​of monitoring indicators.

[0018] Preferably, the root cause location includes log correlation analysis, data correlation analysis and distributed tracing. The log correlation analysis first analyzes the application log, analyzes the error information and abnormal patterns in the log, to determine the error or abnormal root cause, and generates an error link. The analysis method is as follows:

[0019]

[0020] In the formula, For time Error or abnormal service node, is the collection of all service nodes. Indicates time Internal Node The error rate, Representation Node Measurement of impact on the entire system;

[0021] The data correlation analysis compares the relationship between performance indicators and service calls through a correlation analysis formula. The analysis method is as follows:

[0022]

[0023] In the formula, represents the correlation coefficient, is the performance indicator, It is the response time of each node in the service call chain. , Respectively represent The values ​​of performance indicators and response time;

[0024] The distributed tracking serial collector data includes the call link and delay between services. If the service response in a link times out or fails, the call link is retrieved according to the link request ID, and the timestamp, response time and other data of the corresponding link are compared to determine the link that caused the exception.

[0025] Preferably, the performance optimization analyzes the server resource utilization and server performance bottleneck by monitoring the usage of CPU, memory, network and disk resources, combined with the response time and throughput of application services. The server resource utilization analysis method is as follows:

[0026]

[0027] In the formula, For Service The overall resource utilization rate, is the server resource set, For Service In Resources Use the response time below, For Service About Resources utilization rate.

[0028] Preferably, the capacity planning and forecasting analyzes the usage data of CPU, memory and disk resources, combines the business growth curve, and uses a linear regression model or a time series forecasting model to infer future resource requirements. The calculation method is as follows:

[0029]

[0030] In the formula, For historical resource usage data, is the business growth rate, Indicates time The predicted capacity at the time, and are model parameters, and Determined by linear regression models or time series forecasting models.

[0031] Preferably, the operation and maintenance release process monitoring is performed by real-time monitoring of continuous integration and continuous delivery processes. After each code change, the system collects the performance indicators of the service in real time through a collector and calculates the performance impact rate. The calculation method of the performance impact rate is as follows:

[0032]

[0033] In the formula, represents the calculated performance impact rate, and Represents the system performance indicators after and before code deployment, respectively.

[0034] Preferably, the microservice governance further includes service dependency analysis and service health sub-assessment, wherein the service dependency analysis monitors the calling relationship between services in real time, constructs a service dependency graph, and by analyzing the latency, error rate, and throughput of the calling link, the system can identify bottlenecks and abnormal points between services;

[0035] The service health assessment is based on the performance data of each service to assess the service health status and identify performance bottlenecks. The assessment method is as follows:

[0036]

[0037] In the formula, Indicates time Service health score, and Respectively indicate time Indicators of successful service calls and total calls.

[0038] Preferably, the user optimization experience collects client operation behavior data and combines it with server performance data to monitor the full-link user experience.

[0039] Accordingly, the present invention further proposes an observable analysis system based on intelligent analysis, the system being used to implement the above-mentioned observable analysis method, comprising:

[0040] The collector module is used to collect operation and maintenance data and send it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data;

[0041] The observability module performs data analysis based on the operation and maintenance data, including root cause location, performance optimization, capacity planning and prediction, operation and maintenance release process monitoring, microservice governance, and user experience optimization, and sends the results of the data analysis to the output module;

[0042] The output module is used to receive the data analysis results of the observable module and display them to the user.

[0043] Compared with the prior art, the present invention has the following technical effects:

[0044] 1. The analysis method proposed in the present invention realizes comprehensive observability of the system by collecting and analyzing client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data, covering 12 dimensions and more than 18,000 indicators, ensuring full-link monitoring from end users to back-end servers, forming a global system view, and improving system operation efficiency.

[0045] 2. The analysis method proposed in the present invention can observe the root cause location capability, integrate logs, link tracking and performance indicators, and the system can quickly locate the root cause of the fault, reducing the time for manual troubleshooting, and shortening the average fault location time to less than 1 minute.

[0046] 3. The capacity planning and prediction of the analysis method proposed in the present invention avoids over- or under-configuration and improves resource utilization through a capacity prediction model that combines historical data with real-time data.

[0047] 4. The analysis method proposed in the present invention can monitor data in real time through operation and maintenance release process monitoring, automatically feedback the impact of code changes, and greatly reduce the impact of failures caused by automated deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the process of the analysis method of the present invention;

[0049] Figure 2 It is a schematic diagram of the collection process of the collector module according to an embodiment of the present invention;

[0050] Figure 3 It is a schematic diagram of the structure of the analysis system described in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the objectives, technical solutions and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in combination with specific embodiments of the present application and with reference to the accompanying drawings.

[0052] Embodiment 1

[0053] See also Figure 1 , 2 This embodiment is an observable analysis method based on intelligent analysis, including the following steps 1 to 3:

[0054] Step 1: The collector module collects operation and maintenance data and sends it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data.

[0055] In step one, the collector module includes a log collector, a cluster collector and an infrastructure collector.

[0056] The log collector further includes a client log collector that monitors the client's operation behavior and a server log collector that collects the server application log. The client log collector specifically collects the user's operation behavior, such as clicks, errors, delays, etc., generates behavior logs and reports them to Clickhouse for storage and analysis. The server log collector is a log collector deployed in the cluster, responsible for collecting logs generated by the application. The data is transmitted through the Kafka message queue and stored in Clickhouse for subsequent analysis.

[0057] The cluster collector is used to collect cluster performance data and container performance data; specifically, it includes cluster performance collection and container performance collection. Cluster performance collection collects performance indicators in the cluster regularly by configuring cluster connections. The cluster collector dynamically monitors Kubernetes resources and services in the cluster through custom resource definitions (CRDs). CRDs is a mechanism for extending the API provided by Kubernetes, allowing users to define their own resource types and interact with the Kubernetes API server. Container performance collection dynamically adds performance collection targets for containers and services through ServiceMonitor, and aggregates the running status of application services in the container.

[0058] Infrastructure collectors are used to collect server performance data, database performance data, and gateway performance data. Specifically, server performance collection is done by deploying infrastructure collectors on the server to collect CPU, memory and other performance indicators at regular intervals; database performance collection is done by configuring database connection information, authorizing monitor permissions, and collecting database instance CPU, memory, IO, disk and other performance indicators; gateway performance collection is done by configuring gateway connection information to collect gateway throughput, latency, load and other performance data.

[0059] In one embodiment of the present invention, the method for collecting data by the collector module includes active collection and passive reporting. The active collection obtains real-time data through the Prometheus platform deployed in the server, which is suitable for system scenarios requiring high-frequency monitoring to ensure real-time feedback; the passive reporting obtains log data through the Clickhouse database for large-scale log analysis, which can efficiently process large-scale log data and support deep query and historical analysis.

[0060] In some other embodiments of the present invention, the collector module further performs data conversion on the passively reported data, and the conversion method is:

[0061] Extract timestamps and indicator data from the Clickhouse database and convert them into time series data specified by Prometheus;

[0062] Map data fields, map timestamp fields to Prometheus time series timestamps, convert different dimensions or fields to Prometheus labels, and convert numeric fields to values ​​of monitoring indicators;

[0063] Finally, Prometheus is configured to collect this data regularly to ensure the consistency and efficiency of the monitoring system.

[0064] Step 2: The observability module performs data analysis based on the operation and maintenance data. The data analysis includes root cause location, performance optimization, capacity planning and forecasting, operation and maintenance release (DevOps) process monitoring, microservice governance, and user experience optimization.

[0065] In step two, by integrating logs, link tracking and performance indicators, the system can quickly locate the root cause of the fault, reducing the time for manual troubleshooting, and shortening the average fault location time to less than 1 minute. The specific location method is root cause location, which includes log correlation analysis, data correlation analysis and distributed tracking. The log correlation analysis first parses the application log, parses the error information and abnormal patterns in the log to determine the error or abnormal root cause and generate an error link. For example, the response time of a user request suddenly increases. The system identifies the service node where the error occurs and determines its root cause by correlating the upstream and downstream logs. The analysis method of the log correlation method is as follows:

[0066]

[0067] In the formula, For time Error or abnormal service node, is the collection of all service nodes. Indicates time Internal Node The error rate, Representation Node A measure of the impact on the entire system.

[0068] The data correlation analysis compares the relationship between performance indicators and service calls through a correlation analysis formula. The analysis method is as follows:

[0069]

[0070] In the formula, represents the correlation coefficient, is the performance indicator, It is the response time of each node in the service call chain. , Respectively represent The formula is used to analyze the impact of different performance indicators on the response time of each node in the service call chain, thereby providing data support for system performance optimization. For example, when the response time of a Web service increases abnormally, the system first discovers the abnormal CPU usage of the node through the performance monitoring module, and then the distributed tracking module confirms that the upstream service that the service depends on is delayed. The visualization module presents the abnormal service node and call link to the operation and maintenance personnel to help locate the root cause of the fault.

[0071] The distributed tracking series collector data includes the call link and delay between services. If the service response in a link times out or fails, the call link is retrieved according to the link request ID, and the timestamp, response time and other data of the corresponding link are compared to determine the link that caused the abnormality. According to the self-service link analysis, the number of request link calls and request depth are analyzed, and a call graph is constructed to determine whether there is a circular call. For example, service A calls service B, and service B calls service C. Under normal circumstances, the response time is 100ms, but a certain request takes more than 500ms. By tracking the data, it can be found that the response time of service C is abnormal, and the root cause can be attributed to service C.

[0072] By integrating the log, cluster and infrastructure collector functions of the collection module, logs, indicators and link data are collected to achieve observable root cause location. Through log correlation analysis and data correlation analysis, the platform can automatically identify the root cause of the fault and generate a visual call link diagram to assist cross-team collaboration in solving problems.

[0073] On the other hand, the performance optimization analyzes the server resource utilization and server performance bottleneck by monitoring the usage of CPU, memory, network and disk resources, combined with the response time and throughput of application services. The server resource utilization analysis method is as follows:

[0074]

[0075] In the formula, For Service The overall resource utilization rate, is the server resource set, For Service In Resources Use the response time below, For Service About Resources The infrastructure collector can periodically collect the usage of resources such as CPU and memory, and identify service nodes with unbalanced resource utilization by combining the system's business load and the calling relationship between services.

[0076] Performance bottleneck analysis uses the cluster collector to monitor the call link of each service and calculate the response time and throughput of each node. When the response time of a node is significantly higher than the average, the system automatically marks the node as a bottleneck and displays it in the visual interface.

[0077] On the other hand, the capacity planning and forecasting analyzes the usage data of CPU, memory and disk resources, combines the business growth curve, and uses a linear regression model or a time series forecasting model to infer future resource requirements. The calculation method is as follows:

[0078]

[0079] In the formula, For historical resource usage data, is the business growth rate, Indicates time The predicted capacity at the time, and are model parameters, and Determined by linear regression model or time series prediction model. For example, assuming that the resource utilization rate of service A is 60%, the capacity prediction model predicts that the business growth rate will increase by 20% in the next week, and thus it is inferred that service A needs to expand its resources by 20% in advance.

[0080] Using linear regression model to obtain model parameters and The method is as follows:

[0081] Collect sufficient historical resource usage data , such as CPU, memory, network bandwidth and other resources in time usage; collect the business growth rate at the corresponding time point , such as the number of users, transaction volume, traffic and other indicators that reflect business growth; collect actual resource demand data , as the target value of the linear regression model (i.e. the output of the model);

[0082] Perform data preprocessing, such as data cleaning to remove outliers and missing values, data normalization or standardization, so as to accelerate model convergence, and divide the data set into a training set and a test set. In this embodiment, the ratio of 7:3 is used. In other embodiments of the present invention, the training set and the test set may be divided into different ratios according to different requirements;

[0083] Select the linear regression model and construct the formula:

[0084]

[0085] Use the training data set to train the linear regression model. The input of the linear regression model is the feature matrix The output target value is the predicted future resource demand ;

[0086] Select Mean Squared Error (MSE) as the loss function and use gradient descent or least squares method to optimize the parameters and , so that the loss function is minimized;

[0087] Use the test set to verify the performance of the model and select a model whose error between the predicted value and the actual value is less than the set value. and As a parameter for capacity planning and forecasting.

[0088] On the other hand, the operation and maintenance release process monitoring monitors the continuous integration and continuous delivery processes in real time. After each code change, the system collects the performance indicators of the service in real time through the collector and calculates the performance impact rate. The calculation method of the performance impact rate is as follows:

[0089]

[0090] In the formula, represents the calculated performance impact rate, and Represents the system performance indicators after and before code deployment. For example, after a new version of code is deployed, the system's deployment monitoring module detects that the response time of a service has increased by 30%. The feedback mechanism immediately notifies the development and operation and maintenance teams of this anomaly, and the operation and maintenance teams respond quickly and take action to avoid a wide range of user impacts.

[0091] By real-time monitoring of the CI / CD (Continuous Integration / Continuous Delivery) process, the system performance, error rate and user feedback after each code deployment are monitored in real time, helping the development, testing and operation teams to quickly identify online problems caused by deployment operations. The platform improves the team's collaboration efficiency through the feedback mechanism and ensures the stability of the system and the quality of continuous delivery.

[0092] On the other hand, the microservice governance further includes service dependency analysis and service health assessment. The service dependency analysis monitors the calling relationship between services in real time and builds a service dependency graph. By analyzing the latency, error rate and throughput of the calling link, the system can identify bottlenecks and abnormal points between services.

[0093] The service health assessment is based on the performance data of each service to assess the service health status and identify performance bottlenecks. The assessment method is as follows:

[0094]

[0095] In the formula, Indicates time Service health score, and Respectively indicate time Indicators of successful service calls and total calls.

[0096] Microservice governance collects the performance indicators, call links and log data of each microservice in a unified manner through the collection module, providing a comprehensive microservice observability solution. Under the microservice architecture, the platform can uniformly monitor the call relationships and performance data between services, helping the operation and maintenance team to identify service bottlenecks and anomalies in real time.

[0097] The last aspect of this step is that the user optimization experience collects client operation behavior data and combines it with the performance data of the server to monitor the full-link user experience.

[0098] Specifically, by collecting client operation behavior data through the client log collector, the system can monitor the user's operation path, page loading time, click delay and other behavior indicators in real time. When the system detects that the loading time of a page exceeds the expected threshold, it immediately triggers an alarm and locates the problem through data analysis on the server. For example, the page loading time after a user clicks a button increases from 500ms to 2000ms. The system analyzes the page loading log and the request data on the server and finds that the root cause of the problem lies in the delay in the back-end database query. Through the client log collector, the system can monitor the user's interface response time, interface availability and other performance indicators in real time, and can quickly locate client call anomalies.

[0099] Step 3: The observable module sends the results of data analysis to the output module.

[0100] Embodiment 2

[0101] See also Figure 3 This embodiment is an observable analysis system based on intelligent analysis, and the system is used to implement the observable analysis method as described in the first embodiment, including:

[0102] The collector module is used to collect operation and maintenance data and send it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data; the collector module includes a log collector, a cluster collector and an infrastructure collector.

[0103] The log collector further includes a client log collector that monitors the client's operation behavior and a server log collector that collects the server application log. The client log collector specifically collects the user's operation behavior, such as clicks, errors, delays, etc., generates behavior logs and reports them to Clickhouse for storage and analysis. The server log collector is a log collector deployed in the cluster, responsible for collecting logs generated by the application. The data is transmitted through the Kafka message queue and stored in Clickhouse for subsequent analysis.

[0104] The cluster collector is used to collect cluster performance data and container performance data; specifically, it includes cluster performance collection and container performance collection. Cluster performance collection collects performance indicators in the cluster regularly by configuring cluster connections. The cluster collector dynamically monitors Kubernetes resources and services in the cluster through custom resource definitions (CRDs). CRDs is a mechanism for extending the API provided by Kubernetes, allowing users to define their own resource types and interact with the Kubernetes API server. Container performance collection dynamically adds performance collection targets for containers and services through ServiceMonitor, and aggregates the running status of application services in the container.

[0105] Infrastructure collectors are used to collect server performance data, database performance data, and gateway performance data. Specifically, server performance collection is done by deploying infrastructure collectors on the server to collect CPU, memory and other performance indicators at regular intervals; database performance collection is done by configuring database connection information, authorizing monitor permissions, and collecting database instance CPU, memory, IO, disk and other performance indicators; gateway performance collection is done by configuring gateway connection information to collect gateway throughput, latency, load and other performance data.

[0106] The observability module performs data analysis based on the operation and maintenance data, including root cause location, performance optimization, capacity planning and forecasting, operation and maintenance release process monitoring, microservice governance and user experience optimization, and sends the results of the data analysis to the output module.

[0107] The output module is used to receive the data analysis results of the observable module and display them to the user.

[0108] The above is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. An observable analysis method based on intelligent analysis, characterized in that: The following steps are involved: The collector module collects operation and maintenance data and sends it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data; The observability module performs data analysis based on operation and maintenance data, including root cause location, performance optimization, capacity planning and forecasting, operation and maintenance release process monitoring, microservice governance, and user experience optimization; The observability module sends the results of data analysis to the output module; The root cause location includes log correlation analysis, data correlation analysis and distributed tracing. The log correlation analysis first analyzes the application log, analyzes the error information and abnormal patterns in the log, determines the error or abnormal root cause, and generates an error link. The analysis method is as follows: In the formula, For time Error or abnormal service node, is the collection of all service nodes. Indicates time Internal Node The error rate, Representation Node Measurement of impact on the entire system; The data correlation analysis compares the relationship between performance indicators and service calls through a correlation analysis formula. The analysis method is as follows: In the formula, represents the correlation coefficient, is the performance indicator, It is the response time of each node in the service call chain. , Respectively represent The values ​​of performance indicators and response time; The distributed tracking serial collector data includes the call link and delay between services. If the service response in a link times out or fails, the call link is retrieved according to the link request ID, and the timestamp and response time data of the corresponding link are compared to determine the link that caused the exception.

2. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The collector module comprises: The log collector further includes a client log collector that monitors the client operation behavior and a server log collector that collects the server application log; Cluster collector, used to collect cluster performance data and container performance data; Infrastructure collector, used to collect server performance data, database performance data and gateway performance data.

3. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The data collection method of the collector module includes active collection and passive reporting. The active collection obtains real-time data through the Prometheus platform deployed in the server, and the passive reporting obtains log data through the Clickhouse database; The collector module also performs data conversion on the passively reported data, and the conversion method is: Extract timestamps and indicator data from the Clickhouse database and convert them into time series data specified by Prometheus; Map data fields, map timestamp fields to Prometheus time series timestamps, convert different dimensions or fields to Prometheus labels, and convert numeric fields to the values ​​of monitoring indicators.

4. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The performance optimization analyzes server resource utilization and server performance bottlenecks by monitoring the usage of CPU, memory, network and disk resources, combined with the response time and throughput of application services. The server resource utilization analysis method is as follows: In the formula, For Service The overall resource utilization rate, is the server resource set, For Service In Resources Use the response time below, For Service About Resources utilization rate.

5. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The capacity planning and forecasting analyzes the usage data of CPU, memory and disk resources, combines the business growth curve, and uses a linear regression model or a time series forecasting model to infer future resource requirements. The calculation method is as follows: In the formula, For historical resource usage data, is the business growth rate, Indicates time The predicted capacity at the time, and are model parameters, and Determined by linear regression models or time series forecasting models.

6. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The operation and maintenance release process monitoring monitors the continuous integration and continuous delivery processes in real time. After each code change, the system collects the performance indicators of the service in real time through the collector and calculates the performance impact rate. The calculation method of the performance impact rate is as follows: In the formula, represents the calculated performance impact rate, and Represents the system performance indicators after and before code deployment, respectively.

7. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The microservice governance further includes service dependency analysis and service health assessment. The service dependency analysis monitors the calling relationship between services in real time and builds a service dependency graph. By analyzing the latency, error rate and throughput of the calling link, the system can identify bottlenecks and abnormal points between services. The service health assessment is based on the performance data of each service to assess the service health status and identify performance bottlenecks. The assessment method is as follows: In the formula, Indicates time Service health score, and Respectively indicate time Indicators of successful service calls and total calls.

8. The observable analysis method based on intelligent analysis according to claim 1, characterized in that: The user optimization experience collects client operation behavior data and combines it with server performance data to monitor the full-link user experience.

9. An observable analysis system based on intelligent analysis, characterized in that: The system is used to implement the observable analysis method according to any one of claims 1 to 8, comprising: The collector module is used to collect operation and maintenance data and send it to the observable module. The operation and maintenance data includes client logs, server logs, cluster performance data, server performance data, database performance data and gateway performance data; The observability module performs data analysis based on the operation and maintenance data, including root cause location, performance optimization, capacity planning and prediction, operation and maintenance release process monitoring, microservice governance, and user experience optimization, and sends the results of the data analysis to the output module; The output module is used to receive the data analysis results of the observable module and display them to the user.

Images