Data processing method, device, equipment, medium and product for configuration management database

Abstract

This application provides a data processing method, apparatus, device, medium, and product for a configuration management database. It relates to the field of data processing technology. The method includes: acquiring multi-source heterogeneous data, including structured data, semi-structured data, and unstructured data, through a preset data access layer; extracting configuration items and their attribute information from the multi-source heterogeneous data, and constructing a configuration item list based on the configuration items and their attribute information; acquiring multi-source behavioral data, performance data, and business data; generating a dependency graph based on the multi-source behavioral data and the configuration item list using a preset inference model; and predicting various preset target faults based on the configuration item list, dependency graph, performance data, and business data, so as to perform corresponding maintenance operations based on the prediction results. This application uses artificial intelligence technology to drive CMDB upgrades, realizing intelligent fusion of multi-source data, dynamic relationship construction, and predictive maintenance, creating an intelligent decision-making center.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data processing method, apparatus, device, medium and product for configuration management database. Background Technology

[0002] In the field of enterprise IT operations and maintenance management, the Configuration Management Database (CMDB) is a core supporting tool, undertaking the crucial function of recording and managing all IT asset configuration items (CI) and their relationships. With the widespread adoption of cloud computing, microservices, and distributed architectures, enterprise IT environments are becoming increasingly dynamic and complex. Virtual machines and container instances are frequently created and destroyed, service call chains are intertwined and constantly changing, and network traffic and log data exhibit significant multi-source heterogeneous characteristics. This places higher demands on the timeliness, completeness, and intelligent analysis capabilities of the CMDB.

[0003] Currently, the technical solution of the CMDB system mainly revolves around three core aspects. In data collection and processing, configuration item data is obtained through a combination of manual input and periodic scanning tools. These scanning tools can collect basic attributes of resources such as servers and network devices through network probing and API calls. In relationship maintenance, the system relies on manual configuration by operations personnel or the parsing of static configuration files to clarify the hierarchical and dependency relationships between configuration items and generate a topology view. In functional application, based on stored static historical data, it provides data support for basic operations and maintenance tasks such as change management and asset inventory.

[0004] However, traditional CMDBs struggle to efficiently integrate data from multiple sources and formats, including cloud platforms, monitoring systems, and log files, often leading to configuration attribute conflicts and low data quality. Secondly, their dynamic topology construction and update capabilities are insufficient, relying on static rules or manually maintained relational topologies. This makes it impossible to capture dynamic changes in service calls and network dependencies in real time, resulting in a disconnect between the topology and the actual business state. Summary of the Invention

[0005] The data processing method, apparatus, device, medium, and product for configuration management databases provided in this application are used to upgrade CMDB from a static configuration database to an intelligent decision-making center.

[0006] In a first aspect, embodiments of this application provide a data processing method for a configuration management database, the method comprising:

[0007] Multi-source heterogeneous data is acquired through a preset data access layer, including structured data, semi-structured data, and unstructured data.

[0008] The configuration items and their attribute information are extracted from the multi-source heterogeneous data, and a configuration item list is constructed based on the configuration items and their attribute information;

[0009] Acquire multi-source behavioral data, performance data, and business data;

[0010] A dependency graph is generated based on the multi-source behavioral data and the configuration item list using a preset inference model;

[0011] Based on the configuration item list, dependency graph, performance data, and business data, predictions are made for each preset target fault, and corresponding maintenance operations are performed according to the prediction results.

[0012] In one possible implementation, configuration items and their attribute information are extracted from the multi-source heterogeneous data, and a configuration item list is constructed based on the configuration items and their attribute information, including:

[0013] A preset intelligent extraction tool is used to extract configuration items and their attribute information from the multi-source heterogeneous data;

[0014] Generate corresponding identifiers for each configuration item and merge the attribute information of the same configuration item from different data sources to construct a configuration item list;

[0015] In response to conflicting attribute information from different data sources for the same configuration item, the target attribute information is determined based on the data source credibility weighting mechanism, combined with data freshness and historical accuracy.

[0016] In one possible implementation, the multi-source behavioral data includes network layer metadata, application layer call chain data, log interaction data, and component association data; the step of generating a dependency graph based on the multi-source behavioral data and configuration item list using a preset inference model includes:

[0017] A temporal connection graph is constructed by treating configuration items as nodes and the behavioral relationships between configuration items as edges. The edges include temporal features, statistical features, and contextual features.

[0018] A preset inference model is used to determine the clusters of configuration items associated with functions and the implicit dependencies and relationship types between configuration items based on the time-series connection graph;

[0019] Generate cross-layer indirect dependencies based on the configuration item list and preset inference rules;

[0020] Edges corresponding to relationships with confidence scores higher than a preset confidence threshold are stored in the temporal connectivity graph to generate a dependency graph, which is stored in a visual form.

[0021] In one possible implementation, the prediction of each preset target fault based on the configuration item list, dependency graph, performance data, and business data includes:

[0022] By associating the configuration item list, dependency graph, performance data, and business data with the identifiers corresponding to the configuration items, the time-series dataset corresponding to each configuration item can be obtained.

[0023] For each preset target fault, a corresponding prediction model is used to make a prediction based on the time series dataset to obtain the prediction result and confidence interval.

[0024] In one possible implementation, the maintenance operation based on the prediction result includes:

[0025] If the prediction results show that the resource utilization rate of the configuration item will exceed the preset threshold, then calculate the time node to trigger the expansion and generate an expansion work order;

[0026] If the prediction results show that the configuration item has a fault higher than the preset fault threshold, a predictive maintenance work order will be generated and pushed to the operation and maintenance system.

[0027] If multiple related configuration items are predicted to experience resource bottlenecks simultaneously, the root cause of the problem will be located by combining the dependency graph and a root cause analysis report will be output.

[0028] In one possible implementation, the method further includes:

[0029] A data lake is constructed to store multi-source heterogeneous raw data and historical data, so that the data in the data lake can be used for model training and backtracking analysis;

[0030] Message queues are used to process high-throughput real-time data.

[0031] Secondly, embodiments of this application provide a data processing apparatus for a configuration management database, comprising:

[0032] The acquisition module is used to acquire multi-source heterogeneous data through a preset data access layer. The multi-source heterogeneous data includes structured data, semi-structured data, and unstructured data.

[0033] A construction module is used to extract configuration items and their attribute information from the multi-source heterogeneous data, and to construct a configuration item list based on the configuration items and their attribute information;

[0034] The acquisition module is also used to acquire multi-source behavioral data, performance data, and business data;

[0035] The generation module is used to generate a dependency graph based on the multi-source behavioral data and the configuration item list using a preset inference model;

[0036] The prediction module is used to predict each preset target fault based on the configuration item list, dependency graph, performance data, and business data, so as to perform corresponding maintenance operations according to the prediction results.

[0037] Thirdly, embodiments of this application provide a data processing device for a configuration management database, including: a memory and a processor;

[0038] The memory stores computer-executed instructions;

[0039] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0040] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0041] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0042] The data processing method, apparatus, device, medium, and product for configuration management databases provided in this application embodiment pre-set a data access layer as a unified interface to integrate data from different channels and formats, achieving automatic discovery. Accurate extraction of configuration items and their attribute information is crucial; configuration items are the core of configuration management, and attribute information describes their characteristic parameters. Effective extraction algorithms or rules can filter key information from massive amounts of data, improving data relevance and usability. A configuration item list is constructed based on the extracted information, organizing scattered information in an orderly manner for easier subsequent processing, forming reliable and unified CMDB resource data, and solving the problems of inaccurate, incomplete, and outdated data. Different types of data have potential relationships; acquiring multi-source data allows for correlation analysis and the discovery of inherent connections. A pre-set inference model analyzes the dependencies of configuration items based on multi-source behavioral data and the configuration item list, presenting them in a graphical format to intuitively demonstrate mutual influence, making complex relationships clear and easy to understand. Visual modeling helps maintenance personnel understand the system architecture and operating mechanism, providing support for troubleshooting and system optimization. During system operation, dependencies may change; the inference model can dynamically update the graph based on new data, maintaining accuracy and real-time performance. By integrating configuration item lists, dependency graphs, performance and business data, the system predicts preset target faults from multiple perspectives. It also integrates and analyzes multiple data to improve prediction accuracy and reliability, identify risks in advance, and maintain systems based on prediction results, thus achieving a shift from passive response to proactive prevention. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0044] Figure 1An application scenario diagram for a multi-threaded concurrent log reading method provided in this application;

[0045] Figure 2 A flowchart illustrating a data processing method for a configuration management database provided in an embodiment of this application;

[0046] Figure 3 A flowchart illustrating a data processing method for a configuration management database provided in another embodiment of this application;

[0047] Figure 4 A schematic diagram of the structure of a data processing apparatus for a configuration management database provided in an embodiment of this application;

[0048] Figure 5 This is a schematic diagram of the structure of a data processing device for a configuration management database provided in one embodiment of this application.

[0049] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0050] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0051] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0052] To clearly understand the technical solution of this application, the solutions of the prior art will be described in detail first.

[0053] In enterprise-level IT operations and maintenance management systems, configuration management databases (CMDBs) are indispensable core tools, primarily responsible for the overall recording and management of all IT asset configuration items and the relationships between them. With the widespread adoption of cloud computing, microservices, and distributed architectures, the dynamism and complexity of enterprise IT environments have significantly increased. The lifecycles of virtual machines and container instances have shortened considerably, service call chains have become increasingly complex, and network traffic and log data exhibit distinct characteristics of multi-source heterogeneity. This undoubtedly places more stringent demands on the timeliness, completeness, and intelligent analysis capabilities of CMDB data. Currently, the technical implementation of CMDB systems focuses on three core aspects. At the data collection and processing level, configuration item data is primarily obtained through manual entry combined with periodic scanning tools. Scanning tools can utilize network probing, API calls, and other methods to collect basic attribute information of IT resources such as servers and network devices. At the relationship maintenance level, operations and maintenance personnel manually configure or parse static configuration files to clarify the hierarchical and dependency relationships between configuration items, thereby generating a topology view. At the functional application level, based on stored static historical data, data support is provided for basic operations and maintenance tasks such as change management and asset inventory. Nevertheless, traditional CMDBs still have significant shortcomings. First, they struggle to efficiently integrate data from different sources and formats, such as cloud platforms, monitoring systems, and log files, which can easily lead to configuration attribute conflicts and result in poor data quality. Second, they have weak dynamic topology construction and update capabilities, relying on static rules or manually maintained relational topologies, which cannot capture dynamic changes in service calls and network dependencies in real time, causing a serious disconnect between the topology view and the actual business status.

[0054] Therefore, in the face of existing technical problems, to achieve automatic data acquisition and solve the difficulty of integrating multi-source heterogeneous data in traditional CMDBs, a pre-set data access layer is adopted to uniformly manage structured, semi-structured, and unstructured multi-source data. Intelligent extraction logic accurately parses configuration items and their attribute information, constructing a unified configuration item list. This ensures the uniqueness and reliability of CI data while breaking down the format barriers of multi-source data and overcoming the efficiency limitations of data integration. Addressing the issue of static lag in CMDB configuration item relationships and their disconnect from actual business status, a dynamic relationship construction system is established using multi-source behavioral data combined with a pre-set inference model. A time-series connection graph is built with configuration items as nodes and behavioral relationships as edges, automatically identifying configuration item clusters and implicit dependencies, generating a real-time updated dependency graph. This avoids the oversights and lags of manual maintenance from the source, allowing the relationship topology to naturally adapt to business dynamics. To upgrade the CMDB from static recording to predictive decision-making and ensure the forward-looking nature of operational actions, a unified time-series dataset is first formed by linking configuration item lists, dependency graphs, performance data, and business data using unique identifiers for each configuration item. Then, a dedicated predictive model is matched to preset target faults to complete trend analysis and output results. Combined with the prediction results, automated operational operations such as expansion and maintenance are triggered, ensuring the accuracy and timeliness of operational decisions. The entire process relies on data association and model inference to achieve a predictive closed loop, which not only enhances the proactive prevention and control capabilities of operations but also upgrades the CMDB from an information recording center to an intelligent decision-making hub.

[0055] Figure 1 This is an application scenario diagram illustrating the data processing method for the configuration management database provided in this application, such as... Figure 1 As shown in the diagram, the scenario corresponding to the data processing method for the configuration management database provided in this application includes: a server 101, a data processing device 102 for the configuration management database, and a terminal device 103. The data processing device for the configuration management database can be integrated into the server 101.

[0056] Specifically, the data processing device 102 of the configuration management database obtains multi-source heterogeneous data from the server 101 in real time through a preset data access layer, extracts configuration items and their attribute information from the multi-source heterogeneous data, and constructs a configuration item list based on the configuration items and their attribute information. Then, the data processing device 102 of the configuration management database obtains multi-source behavioral data, performance data, and business data from the server 101; and uses a preset inference model to generate a dependency graph based on the multi-source behavioral data and the configuration item list. Finally, the data processing device 102 of the configuration management database predicts various preset target faults based on the configuration item list, dependency graph, performance data, and business data. The prediction results and dependency graph are displayed to the operation and maintenance personnel in a visual form through the terminal device 103.

[0057] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0058] Figure 2 A flowchart illustrating a data processing method for a configuration management database provided in an embodiment of this application is shown below. Figure 2 As shown, the execution entity in this embodiment is a data processing device for a configuration management database. This device can be implemented through a computer program, or through a medium storing the relevant computer program, such as a USB flash drive and / or optical disc, or through a physical device integrating or installing the relevant computer program, such as a chip or a data processing device for a configuration management database. The data processing device for a configuration management database can be a server, a server cluster, etc. The data processing method for a configuration management database provided in this embodiment includes the following steps:

[0059] S201. Obtain multi-source heterogeneous data through a preset data access layer. Multi-source heterogeneous data includes structured data, semi-structured data, and unstructured data.

[0060] Among them, the preset data access layer refers to a standardized data access module designed for receiving and aggregating multi-source heterogeneous data, which integrates a layered architecture of various data adaptation, transmission, and preliminary access rules.

[0061] Among these, "multi-source" refers to data originating from different IT systems and tools within an enterprise, such as cloud platforms, monitoring systems, maintenance ticketing systems, network equipment, and log centers; "heterogeneous" refers to data existing in different structural forms, including structured data, semi-structured data, and unstructured data.

[0062] Structured data includes API interface return data, databases, standardized configuration files, CI attribute tables, etc.

[0063] Semi-structured data includes log files, configuration files, network traffic data, application tracing data, call chain data, etc.

[0064] Unstructured data includes maintenance documents and manuals, work orders, emails, and other information.

[0065] Understandably, before actually acquiring data, it's necessary to complete standardized configuration of the access layer based on the data source types and data transmission requirements of the enterprise's existing IT environment. Dedicated connection channels should be designed for different data sources, such as providing API interface connection capabilities for cloud platforms and monitoring systems, file transfer protocol connection capabilities for log files and configuration files, and application cascading connection capabilities for operation and maintenance ticket systems and email systems. Basic verification standards for data transmission should be pre-defined, including data transmission integrity verification, data source identity and legality verification, and data transmission protocol matching verification, to prevent data loss or tampering during unauthorized data source access or transmission. A lightweight temporary cache space should be built at the access layer to temporarily store massive amounts of real-time data, while pre-setting data routing rules to automatically route incoming data to the corresponding processing channels based on data type, data importance, and transmission frequency. Standardized transmission protocols should be set for all connected data sources to ensure that transmitted data from different data sources can be recognized by the access layer, breaking down transmission protocol barriers between different systems.

[0066] Specifically, structured data such as JSON / XML data returned by cloud platform APIs, database table data in enterprise databases, and standardized indicator data from monitoring systems are obtained through direct API connection retrieval and database synchronous subscription at the access layer. Semi-structured data such as system log files, device configuration files, and service call chain logs are obtained through real-time file collection and streaming data reception at the access layer. Unstructured data such as CMDB maintenance work order descriptions, technical documents, email bodies, and maintenance communication records are obtained through application cascading collection and batch import adaptation at the access layer.

[0067] S202. Extract configuration items and their attribute information from multi-source heterogeneous data, and construct a configuration item list based on the configuration items and their attribute information.

[0068] Configuration items (CI) refer to various manageable resource entities in an enterprise's IT architecture, including servers, databases, network devices, software applications, containers, cloud instances, etc.

[0069] Among them, configuration item attribute information refers to key information describing the characteristics of configuration items.

[0070] The configuration item list refers to a standardized list built based on the extracted configuration items and their attribute information. Each configuration item is assigned a unique identifier, and attribute information from multiple sources is integrated and conflicts are resolved.

[0071] Understandably, data cleaning should be performed before processing the data, including standardization and formatting. This involves resource instances, and special attention should be paid to all timestamps, IP addresses, and MAC address formats, while filtering out outliers.

[0072] Specifically, by utilizing pre-defined mapping rules and field matching logic, configuration items and standardized attributes are directly extracted from structured data. For example, attributes such as virtual machine ID, IP address, and creation time are matched from cloud platform API data. Configuration items and attribute information are parsed from semi-structured data using format parsing and key field recognition technologies. For example, attributes such as hostname, service name, and error code are identified from system logs, and attributes such as service node name, caller, and callee are identified from call chain data. Configuration items and attribute information are mined from text data without a fixed format, relying on NLP and entity recognition technologies. For example, from the maintenance work order "deploy Python 3.9 for my-app-05", the configuration item "server my-app-05" and the attribute "Python 3.9 deployed" are identified.

[0073] Furthermore, a globally unique persistent identifier is generated for each identified configuration item, such as one generated by combining information like IP address, hostname, and device serial number. This identifier is used to determine whether information from different data sources points to the same configuration item. Based on this unique identifier, attribute information for the same configuration item from different data sources is associated and merged. For example, attributes such as the operating system Ubuntu 20.04 provided by the cloud platform API, the IP address 10.0.1.5 provided by the monitoring system, and the deployment of Python 3.9 provided by the maintenance work order are merged into the attribute set of the same configuration item.

[0074] Optionally, when the same attribute of the same configuration item conflicts in different data sources, a decision is made based on the data source credibility weighting mechanism. Credibility weights are pre-defined for different data sources, with weighting reference dimensions including data freshness, historical accuracy, and data source authority. For example, the credibility weight of a cloud platform API is higher than that of a manual support ticket. Conflicting attribute values ​​are compared, and the optimal attribute value is calculated based on the weights. For instance, if the cloud platform API displays the operating system as Ubuntu 20.04, while a log shows CentOS, the cloud platform data is prioritized.

[0075] Furthermore, all configuration items that have completed attribute merging and conflict resolution will be organized into a configuration item list in a unified format. The list includes information such as the unique identifier of the configuration item, type, complete set of attributes, data source, and credibility score, and supports subsequent dynamic updates and expansions.

[0076] S203. Obtain multi-source behavioral data, performance data, and business data.

[0077] Multi-source behavioral data refers to dynamic interactive data generated by enterprise IT resources during operation, reflecting the behavioral states of calls, communication, and collaboration between configuration items.

[0078] Among them, performance data refers to the operational status indicators of various configuration items, which are used to measure the health and resource utilization of configuration items.

[0079] Business data refers to data that is directly related to the enterprise's business processes and reflects the business operation load.

[0080] Specifically, behavioral data is acquired through real-time collection and end-to-end capture, covering multiple dimensions including the network layer, application layer, and container layer. Network layer behavioral data collection involves deploying network traffic collection tools to capture communication data between servers and network devices, extracting information such as source IP, destination IP, communication port, protocol type, traffic volume, and communication frequency. Application layer behavioral data collection involves connecting to a distributed tracing system to capture the call relationships between microservices across the entire chain, extracting link data such as caller, callee, call duration, and request success rate. Application runtime logs are parsed to identify collaborative behaviors between application components. Container / cloud platform behavioral data collection involves connecting to a container management platform to collect the relationships and scheduling behavior data of components such as Pods, Services, and Deployments.

[0081] Furthermore, regarding the operational status of configuration items, multi-dimensional performance metrics are collected through a combination of proactive monitoring and passive reporting. These metrics include hardware performance data, such as CPU utilization, memory usage, disk I / O, disk utilization, and network bandwidth utilization; and software performance data, such as application service QPS, concurrent users, and interface response latency.

[0082] Furthermore, through system integration and data synchronization, it enables the linked analysis of business load and IT performance. It interfaces with the enterprise's core business systems, such as the business middle platform, transaction system, and user management system, to obtain business operation data. It collects business-related operational data from the IT service management system, such as the number of business fault repair work orders, business change request records, and system problem data related to user complaints.

[0083] S204. A dependency graph is generated based on multi-source behavioral data and a list of configuration items using a pre-defined inference model.

[0084] Among them, the preset reasoning model refers to a set of intelligent analysis models that have been pre-trained and configured, mainly including community detection algorithms and graph neural network models, which are used to analyze the behavioral association features of configuration items and reason about the explicit and implicit dependencies between configuration items.

[0085] Among them, the dependency graph refers to a visual topology graph constructed with configuration items as nodes and the dependencies between configuration items as edges. It contains information such as relationship type and confidence level, and can dynamically reflect the real-time association status of IT resources.

[0086] Understandably, after acquiring the data, based on the unique identifier of the configuration item, multi-source behavioral data is mapped to the corresponding configuration item, clearly defining the source and target configuration items for each piece of behavioral data. For example, the source IP and target IP in network traffic data are mapped to two servers in the configuration item list. Multi-dimensional features are extracted from the associated behavioral data, including time-series features, statistical features, and contextual features. Time-series features include communication frequency, call interval, and peak periods; statistical features include the number of interactions and data transmission volume within a specific time period; contextual features include communication protocol type, port number, and API name.

[0087] Specifically, all configuration items in the configuration item list are treated as nodes in a graph, and the behavioral interactions between configuration items are treated as edges connecting the nodes. Each edge is labeled with previously extracted temporal, statistical, and contextual features, forming a temporal connection graph rich in information. A pre-defined inference model is used to analyze the temporal connection graph, uncovering explicit and implicit dependencies between configuration items. A community detection algorithm is applied to analyze the temporal connection graph, automatically identifying groups of configuration items with tightly connected internal components and sparse external connections. These groups typically correspond to a specific business domain or microservice cluster within an enterprise; for example, user services, order services, and payment services are grouped into an e-commerce transaction cluster. Based on a graph neural network model, feature vectors of nodes and edges are learned to infer the dependency types and implicit relationships between configuration items. For example, it is determined that the order service has a read-write dependency with the database and a read-only dependency with the cache; simultaneously, it predicts configuration item relationships that currently have no direct interaction but have potential call requirements. Knowledge graphs and rule-based reasoning methods are used to integrate static configuration and dynamically discovered relationships for logical reasoning. The known static relationships in the CMDB are added to the graph as prior knowledge, and inference rules are defined. The system then automatically infers indirect, cross-layer dependencies based on these rules.

[0088] Furthermore, a confidence score is calculated for each inferred dependency relationship, with scoring dimensions including interaction frequency, feature matching degree, and historical verification accuracy. A confidence threshold is set; relationships exceeding the threshold are automatically included in the graph, while relationships below the threshold are marked as pending review and manually confirmed by operations personnel. Verified configuration item nodes and dependency relationship edges are categorized and displayed according to business domain and relationship type, generating a visual dependency graph. The graph supports hierarchical expansion and relationship tracing, facilitating viewing by operations personnel.

[0089] Optionally, establish real-time update rules for the dependency graph. When new behavioral data or configuration information is collected and changes occur, the model is automatically triggered to re-infer and update the dependency graph to ensure that the graph is consistent with the actual business status.

[0090] S205. Based on the configuration item list, dependency graph, performance data, and business data, predict the faults of each preset target, and perform corresponding maintenance operations according to the prediction results.

[0091] Among them, the preset target faults refer to various types of faults that need to be prevented, which are defined in advance according to the enterprise's IT operation and maintenance needs. These include resource capacity bottleneck faults, hardware faults, service dependency interruption faults, and abnormal business load faults.

[0092] Maintenance operations refer to targeted maintenance actions triggered based on fault prediction results, including proactive maintenance measures such as resource expansion, hardware replacement, service disaster recovery switching, and business rate limiting.

[0093] Specifically, based on the unique identifier of each configuration item, performance data is mapped to the corresponding configuration item, clarifying the IT resource to which each performance metric belongs. Simultaneously, business data is bound to the configuration item cluster that carries the business, for example, associating order volume data with an e-commerce transaction service cluster. The topology and dependency information from the dependency graph are embedded into the prediction dataset to supplement the data on the relationships between configuration items.

[0094] Furthermore, for different types of preset target faults, corresponding prediction models are matched: for resource capacity bottleneck faults, time series prediction models are matched; for hardware faults, survival analysis models are matched; for service dependency interruption faults, association prediction models based on graph neural networks are matched; and for business load anomaly faults, regression analysis models are matched.

[0095] Optionally, corresponding feature data can be extracted from the fused dataset to meet the input requirements of different models. For example, historical time-series data such as disk usage and CPU utilization can be input for time series models; data such as service call frequency and dependency strength can be input for association prediction models. Simultaneously, model parameters can be continuously iterated and optimized based on historical fault data to improve prediction accuracy. Predictive models corresponding to preset target faults can be launched and analyzed in parallel: for example, simultaneously predicting disk capacity bottlenecks of core servers, hardware failure risks of database servers, and dependency interruption risks of order services.

[0096] Furthermore, based on the predicted scope and probability of the failure's impact, the results are categorized into risk levels, such as high, medium, and low. The expected time window for the failure is also output; for example, the disk utilization of server A is expected to exceed the 95% threshold in 15 days. Combining this with a dependency graph, potential related problems caused by the failure are predicted. For instance, if a hardware failure of the core database server is predicted, it is simultaneously predicted that it may lead to the interruption of all services in the e-commerce transaction cluster, generating a complete failure impact analysis report.

[0097] Optionally, based on the risk level and fault type predicted by the forecast, corresponding maintenance operations can be automatically triggered. If a resource capacity bottleneck is predicted, the expansion requirement is automatically calculated and an expansion work order is generated and pushed to the resource management system for execution. If a hardware failure risk is predicted, a hardware replacement work order is automatically generated and the maintenance personnel are notified to prepare replacement parts in advance. If a service dependency interruption risk is predicted, the service disaster recovery switchover process is automatically triggered to switch business traffic to the backup node.

[0098] Optionally, after the maintenance operation is completed, the resource status and business operation are tracked, and the operation effect data is fed back to the prediction model library to optimize model parameters and prediction rules, thereby continuously improving the accuracy of fault prediction and the effectiveness of maintenance operations.

[0099] The data processing method for the configuration management database provided in this application embodiment pre-sets a data access layer as a unified interface to integrate data from different channels and formats, achieving automatic discovery. Accurate extraction of configuration items and their attribute information is crucial; configuration items are the core of configuration management, and attribute information describes their characteristic parameters. Effective extraction algorithms or rules can filter key information from massive amounts of data, improving data relevance and usability. A configuration item list is constructed based on the extracted information, organizing scattered information in an orderly manner for easier subsequent processing, forming reliable and unified CMDB resource data, and solving the problems of inaccurate, incomplete, and outdated data. Different types of data have potential relationships; acquiring multi-source data allows for correlation analysis and uncovering inherent connections. A pre-set inference model analyzes the dependencies of configuration items based on multi-source behavioral data and the configuration item list, presenting them in a graphical format to intuitively demonstrate mutual influence, making complex relationships clear and easy to understand. Visual modeling helps maintenance personnel understand the system architecture and operating mechanism, providing support for troubleshooting and system optimization. During system operation, dependencies may change; the inference model can dynamically update the graph based on new data to maintain accuracy and real-time performance. By integrating configuration item lists, dependency graphs, performance and business data, the system predicts preset target faults from multiple perspectives. It also integrates and analyzes multiple data to improve prediction accuracy and reliability, identify risks in advance, and maintain systems based on prediction results, thus achieving a shift from passive response to proactive prevention.

[0100] As an optional implementation, based on the above embodiments, configuration items and their attribute information are extracted from multi-source heterogeneous data, and a configuration item list is constructed based on the configuration items and their attribute information, including:

[0101] A pre-set intelligent extraction tool is used to extract configuration items and their attribute information from multi-source heterogeneous data;

[0102] Generate corresponding identifiers for each configuration item and merge the attribute information of the same configuration item from different data sources to construct a configuration item list;

[0103] In response to conflicting attribute information from different data sources for the same configuration item, the target attribute information is determined based on the data source credibility weighting mechanism, combined with data freshness and historical accuracy.

[0104] Among them, the preset intelligent extraction tool refers to a set of specialized tools that integrate mapping rules, format parsing, NLP, entity recognition and other capabilities. It adopts differentiated extraction strategies for different types of heterogeneous data to achieve accurate identification and extraction of configuration items and attribute information.

[0105] Among them, the configuration item identifier is a globally unique and persistent identity assigned to each configuration item, which is generated from the core feature information of the configuration item.

[0106] The data source credibility weighting mechanism refers to the pre-defined data source priority evaluation rules, which comprehensively consider the data source's authority, data freshness, historical accuracy, and other dimensions to assign different credibility weight values ​​to different data sources, thereby resolving the attribute conflict problem of the same configuration item.

[0107] Among them, target attribute information refers to the optimal attribute value selected by the credibility weight mechanism when the same attribute of the same configuration item conflicts in different data sources. It is the standard attribute information that is finally written into the configuration item list.

[0108] Specifically, the tool's mapping rules module is invoked to directly match configuration items and attributes in structured data. The tool's format parsing and key field recognition modules are then activated to parse attribute information from semi-structured data. The tool's NLP and entity recognition modules are activated to mine configuration items and attributes from text without a fixed format. After attribute extraction, a preliminary list of configuration items is constructed by assigning unique identifiers and merging attributes from multiple sources. Globally unique identifiers are generated based on the core feature information of the configuration items. Using these identifiers as the matching basis, attribute information for the same configuration item from different data sources is aggregated and merged.

[0109] For example, all attributes provided by the cloud platform API, such as the operating system Ubuntu 20.04, the CPU 8 cores provided by the monitoring system, and the deployment time 2025-01-10 provided by the operation and maintenance work order, are all grouped under the configuration item name corresponding to the identifier.

[0110] Furthermore, all configuration items with merged attributes are organized into a preliminary configuration item list according to a unified format. This list includes basic information such as configuration item identifier, configuration item type, complete attribute set, and data source. When the same attribute of the same configuration item contradicts itself in different data sources, a preset weighting mechanism is used to filter the optimal attribute value and improve the configuration item list. The same attribute field of the same configuration item is compared to identify differing attribute values. For example, if the cloud platform API shows a server operating system as Ubuntu 20.04, while a historical work order shows CentOS 7, this is determined to be an attribute conflict. Preset data source credibility weighting rules are applied, prioritizing data from data sources with more recent updates; reference is made to the accuracy records of the data source in the past, with higher accuracy resulting in higher weight; core system data sources have higher weight than manually entered data sources.

[0111] Furthermore, based on the weight values ​​of each data source, the attribute value with the highest weight is automatically selected as the target attribute information and written into the configuration item list. Finally, all the configuration item information after resolving conflicts is integrated to form a unified and reliable final configuration item list.

[0112] The data processing method for the configuration management database provided in this application uses a preset intelligent extraction tool to extract configuration items and their attribute information. This method can adapt to data from different sources and accurately extract the required information. By merging the attribute information of the same configuration item from various data sources, a complete configuration item view can be constructed, providing a comprehensive understanding of all aspects of the configuration item's characteristics. When the attribute information of the same configuration item conflicts across different data sources, the target attribute information is determined based on a data source reliability weighting mechanism, combined with data freshness and historical accuracy. This comprehensively considers multiple factors and selects the most reliable and up-to-date data. This avoids errors that may result from relying solely on a single data source and improves data accuracy.

[0113] As an optional implementation, based on the above embodiments, the multi-source behavioral data includes network layer metadata, application layer call chain data, log interaction data, and component association data; a dependency graph is generated using a preset inference model based on the multi-source behavioral data and configuration item list, including:

[0114] The configuration items are treated as nodes, and the behavioral relationships between the configuration items are treated as edges to construct a temporal connection graph. The edges include temporal features, statistical features, and contextual features.

[0115] A pre-defined inference model is used to determine the clusters of configuration items related to functions and the implicit dependencies and relationship types between configuration items based on the temporal connection graph;

[0116] Generate cross-layer indirect dependencies based on the configuration item list and preset inference rules;

[0117] Edges corresponding to relationships with confidence scores higher than a preset confidence threshold are stored in a temporal connectivity graph to generate a dependency graph, which is then stored in a visual format.

[0118] Among them, the temporal connection graph refers to a dynamic graph structure built with configuration items as nodes and the behavioral relationships between configuration items as edges. Each edge carries three types of features: temporal, statistical, and contextual, which can reflect the interaction state of configuration items under different time dimensions.

[0119] Among them, the configuration item cluster with functional association refers to a set of configuration items with frequent internal interactions and sparse external interactions. It usually corresponds to a certain business domain or functional module of an enterprise, such as the order service cluster of an e-commerce system.

[0120] Implicit dependencies refer to indirect relationships between configuration items that do not have direct behavioral interactions but exist through business logic or topology.

[0121] Among them, the confidence score is a quantitative indicator that measures the authenticity of a dependency relationship. It is calculated by combining dimensions such as the frequency of behavior of the relationship, the degree of feature matching, and the accuracy of historical verification, and is used to screen reliable dependencies.

[0122] Specifically, all configuration items in the configuration item list are treated as nodes in the time-series connection graph, with each node associated with a unique identifier and core attributes of the configuration item. The configuration item interaction behaviors recorded in the multi-source behavioral data are transformed into edges connecting the nodes. For example, the communication behavior between server A and database B, or the link behavior of service C calling service D, are all treated as edges connecting the corresponding nodes. Temporal features are added to each edge, extracting time-dimensional features from the behavioral data, including interaction frequency, call interval, peak periods, and behavior duration, reflecting the temporal patterns of configuration item interactions. Statistical features are also included, calculating interaction statistics within a specific time window, including total number of interactions, total data transmission, and request success rate. Contextual features are also included, recording environmental information where the behavior occurs, including communication protocol type, call interface name, port number, and business scenario label. Edges with these three types of features are associated with nodes to form the basic time-series connection graph.

[0123] Furthermore, a pre-defined inference model is invoked to analyze the temporal connectivity graph. The community detection algorithm module within this model analyzes the edge density of each node in the graph. Nodes with a large number of internal edges, frequent interactions, and relatively few interactions with external nodes are grouped into a configuration item cluster. The graph neural network module in the model learns the feature vectors of nodes and edges, and identifies implicit dependencies between configuration items through vector similarity calculation and association rule mining. The relationship type is determined based on edge characteristics; for example, active call dependencies and passive response dependencies are distinguished based on the call direction.

[0124] Furthermore, by combining the static attributes of the configuration item list with preset inference rules, cross-level indirect dependencies are derived, enriching the association dimensions of the graph. Predefined inference rules based on IT architecture logic are used; for example, if configuration item A runs on configuration item B, then A indirectly depends on B. The inference is then performed according to preset rules, combining the static attributes of the configuration item list with the discovered direct dependencies.

[0125] For example, given that the order database runs on physical server X and the order service depends on the order database, it can be deduced that the order service indirectly depends on physical server X across layers; if the e-commerce transaction cluster includes the order service, it can be further deduced that the e-commerce transaction cluster indirectly depends on physical server X.

[0126] Furthermore, following a pre-defined scoring formula, a confidence score is calculated for each dependency relationship, considering factors such as edge behavior frequency, feature matching degree, historical verification accuracy, and rule compliance. Dependencies with scores above a threshold are retained in the temporal connectivity graph, while those with scores below the threshold are marked for manual review to prevent invalid relationships from affecting the graph's accuracy. The filtered nodes and edges are then visually arranged according to business domain and relationship type to generate a dependency graph. The graph supports node retrieval, relationship tracing, and hierarchical expansion, allowing operations personnel to intuitively view the association status of configuration items.

[0127] The data processing method for the configuration management database provided in this application uses a graph constructed with configuration items as nodes to intuitively display the existence and distribution of various components in the system. By combining the behavioral relationships between configuration items as edges, the overall architecture and hierarchical structure of the system can be clearly presented. Through inference models, potential dependencies can be revealed, grouping configuration items with similar functions or jointly completing a task together, thereby dividing and analyzing the system from a functional perspective and understanding the collaborative relationships between different functional modules. Based on the configuration item list and preset inference rules, cross-layer indirect dependencies are generated, considering the complex interactions between different layers of the system. By generating cross-layer indirect dependencies, the mutual influence between different layers of the system can be comprehensively evaluated, improving the reliability and stability of the system. By setting reasonable confidence thresholds, the relationships stored in the graph are ensured to have high accuracy and credibility.

[0128] As an optional implementation, based on the above embodiments, prediction of each preset target fault is performed based on a configuration item list, dependency graph, performance data, and business data, including:

[0129] By associating the configuration item list, dependency graph, performance data, and business data with the identifiers corresponding to the configuration items, the time-series dataset corresponding to each configuration item can be obtained.

[0130] For each preset target fault, a corresponding prediction model is used to make predictions based on the time series dataset to obtain the prediction results and confidence intervals.

[0131] Among them, time-series datasets refer to datasets indexed by time that integrate multi-dimensional information such as configuration item attributes, dependencies, performance metrics, and business load, and can reflect the pattern of configuration item status changes over time.

[0132] Specifically, for performance data, collected metrics such as CPU utilization, disk I / O, and memory usage are precisely mapped to corresponding configuration items using configuration item identifiers, forming a performance time-series sequence for each configuration item and marking the performance status at different time points. For business data, business metrics such as order volume, concurrent users, and transaction success rate are bound to the configuration items or configuration item clusters supporting the business based on the business system deployment architecture, clarifying the correspondence between business load and IT resources. Static attribute information from the configuration item list, such as device model, operating system version, and deployment location, is embedded into the dataset as basic information. Relationship information from the dependency graph, such as upstream dependent services and downstream business services of a configuration item, is supplemented into the dataset of the corresponding configuration item, marking the scope of its influence. Using time as a unified index, the static attributes, dynamic performance, business load, and dependency information of each configuration item are integrated to form a time-series dataset specific to each configuration item.

[0133] Understandably, it's advisable to pre-establish rules for matching fault types with predictive models. For example, for resource capacity bottleneck faults, a time series prediction model is used; for hardware aging faults, a survival analysis model is used; for service dependency interruption faults, a graph neural network-based correlation prediction model is used; and for abnormal business load faults, a regression analysis model is used.

[0134] Furthermore, the time-series dataset for each configuration item is processed to extract features according to the input requirements of the corresponding model. For example, trend and periodic features of historical resource utilization are extracted for the time-series model; service call frequency and dependency strength features are extracted for the correlation prediction model. The matching prediction model is then started, and calculations and analyses are performed based on the extracted features to output preliminary prediction results, such as server A's disk utilization will reach the 95% threshold in 15 days, and database B's failure probability is 65% in the next 30 days.

[0135] Furthermore, while outputting prediction results, the model also calculates confidence intervals based on historical prediction accuracy and data volatility. The width of the confidence interval is positively correlated with data stability; the more stable the data, the narrower the confidence interval. The prediction results and confidence intervals are integrated to form a standardized prediction report. This report includes core information such as fault type, affected configuration items, estimated occurrence time, confidence interval, and risk level, providing accurate guidance for subsequent maintenance operations.

[0136] Optionally, based on time series prediction models (such as LSTM), causal reasoning techniques can be introduced, combining dependency graphs and business data mining to explore the causal relationships of resource changes, thereby improving the interpretability and reliability of the prediction results.

[0137] Specifically, based on dependency graphs and business logs, causal graphs are constructed, for example, establishing causal paths between database disk utilization and variables such as user concurrency and service call chain depth. Intervention analysis simulates the impact of changes in external factors on resources; for example, assuming user concurrency doubles, it predicts the growth trend of database disk utilization. The causal graph is then combined with an LSTM model, using causal paths to constrain the LSTM's parameter learning; for example, adjusting the LSTM's sensitivity to user concurrency variables through causal path weights.

[0138] The data processing method for configuration management databases provided in this application, by associating multi-source data to form a time-series dataset, allows the prediction model to acquire more comprehensive and richer information, thereby more accurately capturing the precursors and patterns of fault occurrence and improving the accuracy of fault prediction. By employing corresponding prediction models for each preset target fault, effective monitoring and early warning for different types of faults can be ensured. The prediction results not only provide the probability of fault occurrence but also a confidence interval. The confidence interval reflects the range of uncertainty in the prediction results, providing decision-makers with more comprehensive information.

[0139] As an optional implementation, based on the above embodiments, corresponding maintenance operations are performed according to the prediction results, including:

[0140] If the prediction results show that the resource utilization rate of the configuration item will exceed the preset threshold, then calculate the time node to trigger the expansion and generate an expansion work order;

[0141] If the prediction results show that the configuration item has a fault higher than the preset fault threshold, a predictive maintenance work order will be generated and pushed to the operation and maintenance system.

[0142] If multiple related configuration items are predicted to experience resource bottlenecks simultaneously, the root cause of the problem will be located by combining the dependency graph and a root cause analysis report will be output.

[0143] Among them, the preset threshold refers to the critical value set in advance based on business needs and resource carrying capacity.

[0144] Among them, predictive maintenance work orders refer to preventive operation and maintenance documents generated for high-failure-risk configuration items, which include risk configuration items, failure types, risk levels, and recommended maintenance measures.

[0145] The root cause analysis report refers to a document describing the root cause of a failure when multiple related configuration items simultaneously experience resource bottlenecks, based on dependency graph analysis. It includes information such as the problem phenomenon, the scope of related impacts, the root cause, and solutions.

[0146] Specifically, when the prediction results indicate that the resource utilization rate of a configuration item will exceed a preset threshold, the system uses the predicted resource utilization rate change curve and the preset threshold to calculate the specific time point at which the resource will reach the threshold. A reasonable window for maintenance operations is also reserved; for example, if the disk is predicted to reach 95% utilization in 15 days, the expansion trigger time is set to 12 days later. Based on the historical resource growth rate of the configuration item and the current business load trend, the required expansion resource amount is calculated. A standardized expansion work order is generated, clearly specifying the configuration item identifier, expansion resource type, expansion size, and suggested execution time. The work order is then pushed to the maintenance management system and automatically assigned to the corresponding maintenance team. Simultaneously, the configuration item and work order progress are highlighted on the maintenance dashboard.

[0147] Furthermore, when the prediction results show that the risk of a configuration item failure is higher than the preset failure threshold, the risk is divided into three levels: high, medium, and low, based on the predicted failure probability and the scope of the failure's impact. For example, if the core database server has a failure probability of 70% and affects the entire business system, it is judged as high risk; if the ordinary test server has a failure probability of 60% and only affects the test environment, it is judged as medium risk. Predictive maintenance work orders are generated for high- and medium-risk configuration items. The work order includes information on the risky configuration item, the predicted failure type, the risk level, and recommended maintenance measures. For example, for the risk of hardware aging, it is recommended to replace the hard drive in advance; for the risk of service performance degradation, it is recommended to optimize the configuration or migrate the service. The predictive maintenance work orders are pushed to the operations and maintenance system, and work order priorities are set. High-risk work orders are marked as urgent priority, requiring operations and maintenance personnel to respond within a specified time. Simultaneously, the system tracks the progress of work order processing, and work orders that are not processed on time will automatically trigger reminder notifications.

[0148] Furthermore, when multiple related configuration items are predicted to experience resource bottlenecks simultaneously, the dependency graph is used to locate all configuration items experiencing resource bottlenecks and analyze their direct and indirect dependencies to identify the business clusters or call chains to which these configuration items belong. For example, if the order service, payment service, and order database all experience high CPU usage simultaneously, it can be determined that these three belong to the e-commerce transaction business chain. Reverse tracing is performed using time-series datasets and dependency graphs to analyze the time sequence of resource bottleneck occurrences for each configuration item and determine the bottleneck propagation path. Then, business data is correlated to check for any sudden increases in business volume during the fault window period. Finally, the root cause is comprehensively determined. A root cause analysis report is generated, including the problem symptoms, a list of related configuration items, the bottleneck propagation path, the root cause, and targeted remedial suggestions.

[0149] The data processing method for the configuration management database provided in this application calculates the time node for triggering expansion and generates an expansion work order when the prediction result shows that the resource utilization rate of a configuration item will exceed a preset threshold. This allows for advance planning of expansion matters, ensuring that the expansion operation is completed before resources actually become a bottleneck. If the prediction result shows that a configuration item has a fault higher than a preset fault threshold, a predictive maintenance work order is generated and pushed to the operation and maintenance system, enabling operation and maintenance personnel to take measures for maintenance before the fault actually occurs, greatly improving the stability and reliability of the system. When multiple related configuration items are predicted to experience resource bottlenecks simultaneously, the root cause of the problem is located by combining the dependency relationship graph and a root cause analysis report is output. This precise method of locating the root cause avoids blind troubleshooting and improves the efficiency of problem solving.

[0150] As an optional implementation, based on the above embodiments, it further includes:

[0151] A data lake is constructed to store multi-source heterogeneous raw data and historical data, so that the data in the data lake can be used for model training and backtracking analysis;

[0152] Message queues are used to process high-throughput real-time data.

[0153] A data lake is a massive data storage architecture that centrally stores all of an enterprise's data, capable of storing structured, semi-structured, and unstructured multi-source heterogeneous data in its original format.

[0154] Understandably, this involves building an infrastructure that supports storage of multiple data types, setting a unified data access standard, and being compatible with multi-source heterogeneous data, including structured, semi-structured, and unstructured data. All raw data collected by the pre-defined data access layer, including API interface return data, log files, operation and maintenance work orders, network traffic data, performance monitoring data, etc., are all accessed to the data lake in their raw format without preprocessing, preserving the integrity and original characteristics of the data.

[0155] Specifically, based on the data lifecycle and usage, the data lake is divided into three layers: raw data layer, cleaned data layer, and application data layer. The raw data layer stores all unprocessed raw data, serving as the foundational data assets of the data lake and supporting long-term archiving. The cleaned data layer stores data that has undergone standardization, deduplication, and noise reduction; it is derived from the raw data layer, has a unified format, and can be directly used for model training. The application data layer stores data processed for specific scenarios, such as historical topology data for fault backtracking and time-series indicator data for capacity prediction, supporting rapid querying and analysis. Simultaneously, all data is tagged, including data source, data type, generation time, and associated configuration item identifiers, facilitating subsequent on-demand retrieval and retrieval.

[0156] Furthermore, historical data, including configuration item attribute data, behavioral relationship data, performance indicator data, and historical failure data, is periodically extracted from the cleaned data layer of the data lake and used as training samples to input into the prediction model and relational inference model. By learning the feature patterns in the data, the model parameters are continuously optimized to improve the prediction accuracy and relational inference precision. When it is necessary to trace the causes of historical failures or verify the model's prediction effect, historical data for the corresponding time period is retrieved from the raw or cleaned data layer of the data lake based on conditions such as time range and configuration item identifiers. The IT resource status and data change process at that time are reproduced, and the data differences before and after the failure are compared and analyzed to pinpoint the root cause of the problem; or the model's prediction results are compared with historical actual data to verify the model's accuracy and thus optimize the prediction strategy.

[0157] Optionally, based on real-time data throughput requirements, a message queue supporting high concurrency and low latency can be deployed, with reasonable queue partitioning and storage strategies set to ensure the stability and reliability of data transmission. When high-throughput real-time data is generated, the data producer encapsulates the data into standardized messages and sends them to the message queue, where the message queue temporarily stores and forwards them, thus decoupling the data producer and consumer.

[0158] The data processing method for configuration management databases provided in this application utilizes a data lake, a centralized repository that integrates heterogeneous raw and historical data from different data sources, breaking down data silos and facilitating unified management. The abundant historical data in the data lake provides rich samples for model training, improving model accuracy and generalization ability; historical data also supports backtracking analysis, helping maintenance personnel locate fault causes and optimize the system. Message queues, acting as middleware, decouple data producers and consumers in real-time data processing, preventing data loss or system congestion. Employing a distributed architecture and asynchronous processing mechanism, it can process messages in parallel, supporting high-throughput real-time data streams and ensuring real-time performance and accuracy.

[0159] Figure 3 A flowchart illustrating a data processing method for a configuration management database provided in another embodiment of this application is shown below. Figure 3 As shown, the data processing method for the configuration management database provided in this embodiment includes the following steps:

[0160] S301. Obtain multi-source heterogeneous data through a preset data access layer. Multi-source heterogeneous data includes structured data, semi-structured data, and unstructured data.

[0161] S302. Use a preset intelligent extraction tool to extract configuration items and their attribute information from multi-source heterogeneous data.

[0162] S303. Generate corresponding identifiers for each configuration item and merge the attribute information of the same configuration item from different data sources to construct a configuration item list.

[0163] S304. In response to conflicting attribute information from different data sources for the same configuration item, the target attribute information is determined based on the data source credibility weighting mechanism, combined with data freshness and historical accuracy.

[0164] S305: Acquire multi-source behavioral data, performance data, and business data.

[0165] S306. Construct a temporal connection graph by treating configuration items as nodes and the behavioral relationships between configuration items as edges. The edges include temporal features, statistical features, and contextual features.

[0166] S307. Using a preset inference model, determine the clusters of configuration items related to functions and the implicit dependencies and relationship types between configuration items based on the time sequence connection graph.

[0167] S308. Generate cross-layer indirect dependencies based on the configuration item list and preset inference rules.

[0168] S309. Store the edges corresponding to relationships with confidence scores higher than the preset confidence threshold in the temporal connection graph to generate a dependency graph, which is then stored in a visual form.

[0169] S310. Associate the configuration item list, dependency graph, performance data, and business data with the identifiers corresponding to the configuration items to obtain the time-series dataset corresponding to each configuration item.

[0170] S311. For each preset target fault, a corresponding prediction model is used to make a prediction based on the time series dataset to obtain the prediction results and confidence intervals.

[0171] S312. If the prediction result shows that the resource utilization rate of the configuration item will exceed the preset threshold, calculate the time node to trigger the expansion and generate an expansion work order.

[0172] S313. If the prediction result shows that the configuration item has a fault higher than the preset fault threshold, a predictive maintenance work order is generated and pushed to the operation and maintenance system.

[0173] S314. If it is predicted that multiple related configuration items will experience resource bottlenecks at the same time, the root cause of the problem will be located by combining the dependency graph and a root cause analysis report will be output.

[0174] It should be noted that the execution order of S312-S314 is not important.

[0175] In this embodiment, the implementation method and technical effect of S301-S314 are similar to those of the corresponding solutions in the above embodiments, and will not be repeated here.

[0176] Figure 4 A schematic diagram of the data processing device for the configuration management database provided in this application is shown below. Figure 4 As shown, the data processing device 40 for the configuration management database provided in this embodiment includes: an acquisition module 41, a construction module 42, a generation module 43, and a prediction module 44.

[0177] The system includes: an acquisition module 41, which acquires multi-source heterogeneous data through a preset data access layer, including structured data, semi-structured data, and unstructured data; a construction module 42, which extracts configuration items and their attribute information from the multi-source heterogeneous data and constructs a configuration item list based on the configuration items and their attribute information; an acquisition module 41, which also acquires multi-source behavioral data, performance data, and business data; a generation module 43, which uses a preset inference model to generate a dependency graph based on the multi-source behavioral data and the configuration item list; and a prediction module 44, which predicts the faults of each preset target based on the configuration item list, dependency graph, performance data, and business data, and performs corresponding maintenance operations based on the prediction results.

[0178] The data processing device for the configuration management database provided in this embodiment can execute... Figure 2 The implementation principles and technical effects of the methods shown are similar, and will not be repeated here.

[0179] Optionally, the construction module 42, when extracting configuration items and their attribute information from multi-source heterogeneous data and constructing a configuration item list based on the configuration items and their attribute information, specifically performs the following: using a preset intelligent extraction tool to extract configuration items and their attribute information from multi-source heterogeneous data; generating corresponding identifiers for each configuration item and merging the attribute information of the same configuration item from different data sources to construct a configuration item list; and, in response to conflicts in the attribute information of the same configuration item from different data sources, determining the target attribute information based on the data source credibility weight mechanism, combined with data freshness and historical accuracy.

[0180] Optionally, the multi-source behavioral data includes network layer metadata, application layer call chain data, log interaction data, and component association data; the generation module 43, when generating a dependency graph based on the multi-source behavioral data and configuration item list using a preset inference model, is specifically used to: construct a temporal connection graph with configuration items as nodes and behavioral relationships between configuration items as edges, the edges including temporal features, statistical features, and contextual features; determine the cluster of configuration items with functional associations and the implicit dependencies and relationship types between configuration items based on the temporal connection graph using a preset inference model; generate cross-layer indirect dependencies based on the configuration item list and preset inference rules; and store the edges corresponding to relationships with confidence scores higher than a preset confidence threshold in the temporal connection graph to generate a dependency graph, which is stored in a visual form.

[0181] Optionally, when predicting each preset target fault based on the configuration item list, dependency graph, performance data, and business data, the prediction module 44 is specifically used to: associate the configuration item list, dependency graph, performance data, and business data with the identifier corresponding to the configuration item to obtain the time series dataset corresponding to each configuration item; and use the corresponding prediction model to predict each preset target fault based on the time series dataset to obtain the prediction result and confidence interval.

[0182] Optionally, the data processing device for the configuration management database provided in this embodiment further includes a computing module.

[0183] Correspondingly, the calculation module is used to calculate the time node for triggering expansion and generate an expansion work order if the prediction result shows that the resource utilization rate of the configuration item will exceed the preset threshold; the generation module 43 is also used to generate a predictive maintenance work order and push it to the operation and maintenance system if the prediction result shows that the configuration item has a fault higher than the preset fault threshold; if it is predicted that multiple related configuration items will have resource bottlenecks at the same time, the root cause of the problem will be located by combining the dependency relationship graph and the root cause analysis report will be output.

[0184] Optionally, the data processing device for the configuration management database provided in this embodiment further includes a processing module.

[0185] Correspondingly, the construction module 42 is also used to construct a data lake to store multi-source heterogeneous raw data and historical data, so as to use the data in the data lake for model training and backtracking analysis; the processing module is used to process high-throughput real-time data using message queues.

[0186] Figure 5 A schematic diagram of the data processing device for the configuration management database provided in this application. (See attached diagram.) Figure 5 As shown, the data processing device 50 for the configuration management database provided in this embodiment includes a processor 51 and a memory 52. ​​The processor 51 and the memory 52 are connected via a bus and communicate with each other.

[0187] In the specific implementation process, the processor 51 executes the computer execution instructions stored in the memory 52, causing the processor 51 to perform the above-described method.

[0188] The specific implementation process of processor 51 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0189] In the above embodiments, it should be understood that the processor 51 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0190] The memory 52 may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0191] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0192] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0193] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0194] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0195] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0196] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0197] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0199] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0200] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0201] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A data processing method for a configuration management database, characterized in that, The method includes: Multi-source heterogeneous data is acquired through a preset data access layer, including structured data, semi-structured data, and unstructured data. The configuration items and their attribute information are extracted from the multi-source heterogeneous data, and a configuration item list is constructed based on the configuration items and their attribute information; Acquire multi-source behavioral data, performance data, and business data; A dependency graph is generated based on the multi-source behavioral data and the configuration item list using a preset inference model; Based on the configuration item list, dependency graph, performance data, and business data, predictions are made for each preset target fault, and corresponding maintenance operations are performed according to the prediction results.

2. The method according to claim 1, characterized in that, The step of extracting configuration items and their attribute information from the multi-source heterogeneous data, and constructing a configuration item list based on the configuration items and their attribute information, includes: A preset intelligent extraction tool is used to extract configuration items and their attribute information from the multi-source heterogeneous data; Generate corresponding identifiers for each configuration item and merge the attribute information of the same configuration item from different data sources to construct a configuration item list; In response to conflicting attribute information from different data sources for the same configuration item, the target attribute information is determined based on the data source credibility weighting mechanism, combined with data freshness and historical accuracy.

3. The method according to claim 1, characterized in that, The multi-source behavioral data includes network layer metadata, application layer call chain data, log interaction data, and component association data; the generation of a dependency graph based on the multi-source behavioral data and configuration item list using a preset inference model includes: A temporal connection graph is constructed by treating configuration items as nodes and the behavioral relationships between configuration items as edges. The edges include temporal features, statistical features, and contextual features. A preset inference model is used to determine the clusters of configuration items associated with functions and the implicit dependencies and relationship types between configuration items based on the time-series connection graph; Generate cross-layer indirect dependencies based on the configuration item list and preset inference rules; Edges corresponding to relationships with confidence scores higher than a preset confidence threshold are stored in the temporal connectivity graph to generate a dependency graph, which is stored in a visual form.

4. The method according to claim 1, characterized in that, The prediction of each preset target fault based on the configuration item list, dependency graph, performance data, and business data includes: By associating the configuration item list, dependency graph, performance data, and business data with the identifiers corresponding to the configuration items, the time-series dataset corresponding to each configuration item can be obtained. For each preset target fault, a corresponding prediction model is used to make a prediction based on the time series dataset to obtain the prediction result and confidence interval.

5. The method according to claim 4, characterized in that, The maintenance operations based on the prediction results include: If the prediction results show that the resource utilization rate of the configuration item will exceed the preset threshold, then calculate the time point for triggering expansion and generate an expansion work order; If the prediction results show that the configuration item has a fault higher than the preset fault threshold, a predictive maintenance work order will be generated and pushed to the operation and maintenance system. If multiple related configuration items are predicted to experience resource bottlenecks simultaneously, the root cause of the problem will be located by combining the dependency graph and a root cause analysis report will be output.

6. The method according to claim 1, characterized in that, The method further includes: A data lake is constructed to store multi-source heterogeneous raw data and historical data, so that the data in the data lake can be used for model training and backtracking analysis; Message queues are used to process high-throughput real-time data.

7. A data processing apparatus for configuring and managing a database, characterized in that, include: The acquisition module is used to acquire multi-source heterogeneous data through a preset data access layer. The multi-source heterogeneous data includes structured data, semi-structured data, and unstructured data. A construction module is used to extract configuration items and their attribute information from the multi-source heterogeneous data, and to construct a configuration item list based on the configuration items and their attribute information; The acquisition module is also used to acquire multi-source behavioral data, performance data, and business data; The generation module is used to generate a dependency graph based on the multi-source behavioral data and the configuration item list using a preset inference model; The prediction module is used to predict each preset target fault based on the configuration item list, dependency graph, performance data, and business data, so as to perform corresponding maintenance operations according to the prediction results.

8. A data processing device for configuring and managing a database, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images