A server full-life-cycle intelligent operation and maintenance method and system based on digital twinning

Abstract

The application discloses a kind of server full life cycle intelligent operation and maintenance methods based on digital twinning, by constructing the digital twin model containing physical layer, logic layer and business layer mapping, and according to load index distinguishes idle time and busy time state;Idle time injects load and fault event to carry out offline simulation, constructs the state database containing normal / abnormal mode and generates dynamic decision criteria and candidate strategy set;Busy time online compares real-time data and state database, according to dynamic criteria alarm, after matching strategy, in digital twin simulation verification and issue execution;Full life cycle data is uniformly stored, and the model, criteria and strategy are continuously iterated and optimized.The application overcomes the defects of existing operation and maintenance stage fragmentation, fixed threshold and lack of verification, realizes idle busy division of closed-loop intelligent operation and maintenance, significantly improves alarm accuracy, strategy safety and operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for data centers, specifically to an intelligent operation and maintenance method and system for the entire lifecycle of servers based on digital twins. Background Technology

[0002] Digital twin technology, as a key enabling technology for the deep integration of the physical and information worlds, has been widely applied in recent years in fields such as industrial manufacturing, smart cities, and aerospace. In the field of data center operations and maintenance management, with the continuous expansion of server scale and the increasing complexity of system architecture, traditional operations and maintenance management methods are facing severe challenges.

[0003] Server operation and maintenance management refers to the entire lifecycle management process of server equipment in a data center, from deployment and online operation, operation monitoring, fault handling to retirement and decommissioning. It encompasses multiple stages such as resource allocation, status monitoring, fault diagnosis, and policy optimization. Currently, the industry commonly adopts a combination of static configuration verification tools based on preset rules and a human experience-driven operation and maintenance management model. Specifically, during the server deployment phase, operation and maintenance personnel import static data such as server hardware parameters, power capacity, and port count, and perform conflict screening based on fixed thresholds to generate compliance reports. During the operation and maintenance phase, performance indicators are collected through monitoring systems, and operation and maintenance personnel determine anomalies and formulate handling plans based on threshold rules.

[0004] Chinese invention patent CN121173691B discloses a server operation and maintenance simulation method and system based on digital twins, belonging to the field of digital twin technology. This method collects deployment configuration data of the server to be deployed, converts physical location information into logical coordinates and forms a resource dependency graph, constructs an interactive 3D visualized topology model in the digital twin environment, detects resource configuration conflicts by injecting simulated runtime load, and verifies the feasibility of the deployment plan by simulating the server online process. Finally, it generates an optimized server allocation table and network configuration strategy. This solution utilizes digital twin technology to achieve simulation verification and optimization during the server deployment phase, effectively improving deployment efficiency and reliability.

[0005] However, the applicant's research revealed that this solution only utilizes static configuration data to build a digital twin model and conduct simulation verification during the server deployment phase. After the server is officially launched, it faces the following fundamental technical obstacles, preventing the digital twin model from being extended to the operation and maintenance phase and making it difficult to achieve closed-loop management throughout its entire lifecycle:

[0006] First, when the server cluster is actually running business, it generates a massive amount of high-frequency indicators per second, covering thousands of dimensions such as CPU microarchitecture status, memory paging, disk I / O, network packet sending and receiving, and application call chain. If it is necessary to maintain the millisecond-level state synchronization between the digital twin and the physical cluster, it will cause huge overhead on network bandwidth and computing resources, which can easily lead to transmission delays and model update lags, causing the twin state to gradually deviate from the real system.

[0007] Second, the lack of traditional digital twin maintenance requires the continuous use of considerable computing power for state synchronization and calculation. During peak business periods, it competes with production tasks for resources. If full synchronization is forced, it will slow down business response. If the speed is limited, it will further aggravate model distortion. There is an irreconcilable conflict between the two.

[0008] Third, factors such as hardware aging, firmware updates, and dynamic changes in microservice topology constantly alter the behavior of the system during server operation. As the simulation deviation of a digital twin model built solely based on the initial configuration increases over time, it becomes difficult to accurately reproduce fault scenarios and verify operation and maintenance strategies.

[0009] Fourth, the experience accumulated during operation in terms of state changes, abnormal events, and handling has not been linked with the digital twin model, and cannot be extracted into a reusable operation and maintenance knowledge base to guide future decision-making.

[0010] The above reasons together lead to a regression to the traditional operation and maintenance model that relies on fixed thresholds and human experience after deployment. This results in low accuracy of fault warnings, delayed policy response, and the inability to pass on operation and maintenance knowledge, which fails to meet the intelligent operation and maintenance needs of large-scale data centers that cover the entire lifecycle. Summary of the Invention

[0011] To overcome the above-mentioned technical deficiencies, this invention provides a server full lifecycle intelligent operation and maintenance method and system based on digital twins.

[0012] To solve the above problems, the present invention is implemented according to the following technical solution:

[0013] In a first aspect, the present invention provides a server full lifecycle intelligent operation and maintenance method based on digital twins, including:

[0014] Obtain the physical configuration information and running status information of the target server cluster, construct a digital twin model corresponding to the target server cluster, the digital twin model includes digital representations of at least one of the physical layer, logical layer and business layer, and establish mapping relationships between the representation layers;

[0015] Detect the load index of the target server cluster, and determine whether the current operating state is an idle state or a busy state based on the load index;

[0016] When the system is determined to be in an idle state, an offline simulation and knowledge base construction process is executed: a preset load dataset and fault event set are input into the digital twin model to simulate the evolution of resource and business states and generate simulation analysis results; a state database is constructed and updated based on the simulation analysis results, which stores normal state patterns, abnormal state characteristics, and performance baselines under different operating conditions; and dynamic judgment criteria and candidate operation and maintenance strategy sets are generated and optimized based on the simulation analysis results and historical operating data.

[0017] When a busy state is detected, an online monitoring and real-time decision-making process is executed: real-time operating status data of the target server cluster is acquired, the real-time operating status data is compared with the normal state mode in the status database, and an alarm signal is generated according to the current dynamic judgment criteria; in response to the alarm signal, a target operation and maintenance strategy is matched from the candidate operation and maintenance strategy set, and the target operation and maintenance strategy is simulated and verified in the digital twin model; the verified target operation and maintenance strategy is sent to the target server cluster for execution, and the execution result information is obtained;

[0018] The full lifecycle status information, log information, alarm information, and operation and maintenance strategy execution information obtained during the idle and busy periods are stored in the status database.

[0019] Using the information in the state database, the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set are iteratively updated.

[0020] In conjunction with the first aspect, the present invention provides a first specific implementation of the first aspect. Specifically, the step of obtaining the physical configuration information and operating status information of the target server cluster, constructing a digital twin model corresponding to the target server cluster, wherein the digital twin model includes digital representations of at least one of the physical layer, logical layer, and business layer, and establishing mapping relationships between the representation layers, specifically includes:

[0021] Hardware configuration information of heterogeneous devices in the target server cluster is collected through a standardized interface. The hardware configuration information includes at least device identifier, performance parameters and physical location information.

[0022] Based on the hardware configuration information, a digital representation of the physical layer is generated, and the digital representation of the physical layer includes logical mapping nodes for each of the heterogeneous devices.

[0023] Identify the physical connection relationships and business call relationships between each of the logical mapping nodes, and construct a resource dependency graph representing the association between resources as a digital representation of the logical layer;

[0024] Based on the nodes and relationships in the resource dependency graph, a visualization representation framework for the digital twin model is generated, which serves as the digital representation of the business layer.

[0025] Establish the mapping relationship between the logical mapping node, the resource dependency graph and the visualization representation framework to form an association mapping between the physical layer, the logical layer and the business layer.

[0026] In conjunction with the first aspect, the present invention provides a second specific implementation of the first aspect. Specifically, the step of inputting a preset load dataset and fault event set into the digital twin model, simulating the evolution process of resource status and business status in the digital twin model, and generating simulation analysis results specifically includes:

[0027] Obtain preset load policies and fault policies, generate corresponding load datasets and fault event sets, and inject the load datasets and fault event sets into the digital twin model. The load policies include business load simulation policies and deployment and online simulation policies, and the fault policies include hardware fault injection policies and network fault injection policies.

[0028] The digital twin model is driven to run a simulation based on the injected data, and the resource status change sequence and service call status change sequence of each node are recorded during the simulation. The resource status change sequence includes at least resource consumption trend, resource conflict marker and conflict node information, and the service call status change sequence includes at least service call path change and cascading impact range.

[0029] The simulation analysis results are generated based on the resource status change sequence and the service call status change sequence. The simulation analysis results include at least one of resource bottleneck analysis, capacity prediction, fault impact range assessment, or deployment scheme verification.

[0030] In conjunction with the first aspect, the present invention provides a third specific implementation of the first aspect. Specifically, the step of comparing the real-time operating status data with the normal status patterns in the status database and generating an alarm signal based on the current dynamic judgment criteria specifically includes:

[0031] The collected real-time operating status data is synchronized to the digital twin model to drive the digital twin model to perform real-time monitoring of multi-dimensional performance indicators of the target server cluster.

[0032] Historical normal state patterns, indicator trend characteristics, and periodic characteristics are extracted from the state database and combined with the current real-time operating data to generate dynamic alarm thresholds and anomaly detection models, which serve as the dynamic judgment criteria.

[0033] The real-time monitoring data of the multi-dimensional performance indicators are input into the dynamic judgment criteria to determine whether the alarm triggering conditions are met, and an alarm signal carrying the alarm level and impact range is generated when the conditions are met.

[0034] In conjunction with the first aspect, the present invention provides a fourth specific implementation of the first aspect. Specifically, the step of matching a target operation and maintenance strategy from the candidate operation and maintenance strategy set in response to the alarm signal, and performing simulation verification of the target operation and maintenance strategy in the digital twin model, specifically includes:

[0035] Extract key feature information from the alarm signal, including at least the fault type identifier, the range of affected resources, and the degree of performance degradation;

[0036] The key feature information is matched with a preset candidate operation and maintenance strategy library to obtain at least one candidate operation and maintenance strategy, and the target operation and maintenance strategy is determined from it according to the preset matching degree sorting rules.

[0037] The target operation and maintenance strategy is simulated and executed in the digital twin model to obtain simulation execution results, which include changes in resource status and service recovery status after execution.

[0038] The simulation execution results are compared with preset verification conditions, and the target operation and maintenance strategy is determined to pass the verification based on the comparison results.

[0039] In conjunction with the first aspect, the present invention provides a fifth specific implementation of the first aspect. Specifically, the step of distributing the verified target operation and maintenance strategy to the target server cluster for execution and obtaining execution result information specifically includes:

[0040] The verified target operation and maintenance strategy is converted into an operation instruction set that can be recognized by the target server cluster, and the operation instruction set is sent to the target server cluster through a standardized interface;

[0041] During the execution process, the real-time running status data of the target server cluster is continuously acquired and compared in real time with the simulated execution results in the digital twin model to monitor whether the execution process deviates from expectations.

[0042] After the operation instruction set is executed, the actual running status data of the target server cluster is obtained as the execution result information. The execution result information includes at least resource status changes, business recovery status, and execution time.

[0043] In conjunction with the first aspect, the present invention provides a sixth specific implementation of the first aspect. Specifically, storing the full lifecycle status information, log information, alarm information, and operation and maintenance strategy execution information obtained in the idle and busy states into the status database specifically includes:

[0044] Acquire multi-source heterogeneous operation and maintenance data generated during idle and busy periods. The multi-source heterogeneous operation and maintenance data includes at least configuration status information, operation log information, historical alarm information, and operation and maintenance strategy execution information.

[0045] Add timestamps and event association tags to the multi-source heterogeneous operation and maintenance data to establish temporal and causal relationships between the operation and maintenance data, forming a structured operation and maintenance event chain;

[0046] Based on the operation and maintenance event chain, operation and maintenance scenario features, fault root cause features, and strategy effectiveness evaluation indicators are extracted to generate an operation and maintenance knowledge structure containing feature tags and retrieval indexes, which is then stored in the status database.

[0047] Secondly, the present invention also provides a server full lifecycle intelligent operation and maintenance system based on digital twins, comprising:

[0048] The model building module is used to obtain the physical configuration information and running status information of the target server cluster, construct a digital twin model corresponding to the target server cluster, the digital twin model includes digital representation of at least one of the physical layer, logical layer and business layer, and establishes the mapping relationship between the representation layers;

[0049] The load status judgment module is used to detect the load index of the target server cluster and determine whether the current running status is an idle state or a busy state based on the load index.

[0050] The idle time processing module is used to perform offline simulation and knowledge base construction when the state is determined to be idle: inputting a preset load dataset and fault event set into the digital twin model, simulating the evolution of resource status and business status, and generating simulation analysis results; constructing and updating a status database based on the simulation analysis results, the status database storing normal status patterns, abnormal status characteristics and performance baselines under different operating conditions; and generating and optimizing dynamic judgment criteria and candidate operation and maintenance strategy sets according to the simulation analysis results and historical operation data.

[0051] The busy-hour processing module is used to perform online monitoring and real-time decision-making processes when a busy-hour state is determined: acquiring real-time operating status data of the target server cluster, comparing the real-time operating status data with the normal state mode in the status database, and generating an alarm signal based on the current dynamic judgment criteria; responding to the alarm signal, matching a target operation and maintenance strategy from the candidate operation and maintenance strategy set, and simulating and verifying the target operation and maintenance strategy in the digital twin model; and sending the verified target operation and maintenance strategy to the target server cluster for execution and obtaining execution result information.

[0052] The information storage module is used to store the full lifecycle status information, log information, alarm information and operation and maintenance strategy execution information obtained in the idle and busy states into the status database;

[0053] The iterative update module is used to iteratively update the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set using information from the state database.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] By detecting load metrics to distinguish between idle and busy periods, offline simulation and database construction are performed during idle periods, while online monitoring and real-time decision-making are performed during busy periods. The entire lifecycle information of both stages is uniformly stored and iteratively updated, which solves the problem that existing solutions only focus on a single stage and do not utilize idle resources for extrapolation. This systematically improves the operation and maintenance efficiency and intelligence level of the server throughout its entire lifecycle.

[0056] By establishing digital representations covering at least one of the physical, logical, and business layers and their inter-layer mapping relationships, not only is a comprehensive mirroring of server hardware, logical dependencies, and business status achieved, but also the multi-layered evolution process from underlying resource anomalies to upper-layer business impacts can be accurately simulated. This provides a high-fidelity simulation foundation for offline simulation and online verification, overcoming the shortcomings of existing technologies that only visualize topology at the deployment stage and cannot extend knowledge to the operation and maintenance stage.

[0057] During idle periods, offline simulation is performed by injecting a preset load dataset and fault event set to build a state database containing normal state modes, abnormal state characteristics, and performance baselines. Dynamic judgment criteria are generated based on simulation and historical data, which enables alarm thresholds to be adaptively adjusted according to factors such as equipment aging and load changes. This effectively solves the problems of missed alarms and false alarms caused by fixed thresholds and significantly improves the accuracy and timeliness of alarms during busy periods.

[0058] During idle-time simulations, a set of candidate operation and maintenance strategies is automatically generated and optimized. When an alarm is generated during busy-time, the target strategy is matched from the strategy set and first verified in a digital twin. Only strategies that pass the verification are then deployed to the physical cluster for execution. This mechanism changes the current situation of relying on human experience to formulate strategies, eliminates the security risks that may be introduced by direct strategy execution, and ensures the effectiveness and reliability of operation and maintenance operations. Attached Figure Description

[0059] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0060] Figure 1 This is a flowchart of an intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins, according to the present invention.

[0061] Figure 2 This is a block diagram of an electronic device used to implement embodiments of the present invention;

[0062] In the picture:

[0063] 100 - Electronic device, 101 - Computing unit, 102 - ROM, 103 - RAM, 104 - Bus, 105 - I / O interface, 106 - Input unit, 107 - Output unit, 108 - Storage unit, 109 - Communication unit. Detailed Implementation

[0064] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0065] like Figures 1-2 As shown, this invention provides a server full lifecycle intelligent operation and maintenance method and system based on digital twins.

[0066] Example 1

[0067] like Figure 1 As shown in the diagram, this invention provides a flowchart of a server intelligent operation and maintenance method based on digital twins throughout its entire lifecycle. This invention provides a server intelligent operation and maintenance technology that constructs a digital twin model of the target server cluster, distinguishes between idle and busy states based on real-time load indicators, performs offline simulation and knowledge construction during idle states, and performs online monitoring and real-time decision-making during busy states, thereby achieving intelligent operation and maintenance throughout the entire lifecycle of the server cluster. This method can handle the operation and maintenance needs under two different operating scenarios, and is particularly suitable for server operation and maintenance management in large-scale data centers, cloud computing platforms, and complex business environments. It reduces the workload of manual inspection and passive response in traditional operation and maintenance, and improves operation and maintenance efficiency and fault handling accuracy.

[0068] The data processing method of this invention can be executed by a server intelligent operation and maintenance system. This system can be implemented in hardware and / or software and can be configured in electronic devices of a data center management platform or cloud management platform, such as an operation and maintenance management server or a backend data processing server. The data processing method of this invention is based on digital twin technology. Specifically, the digital twin system includes a data acquisition module, a digital twin engine module, a simulation analysis module, a policy matching module, and a policy distribution module, capable of quickly acquiring the real-time status of the server cluster and performing simulation analysis. Figure 1 As shown, the data processing method includes:

[0069] S100: Obtain the physical configuration information and initial operating status information of the target server cluster, and construct a digital twin model corresponding to the target server cluster.

[0070] The digital twin model includes digital representations of at least one of the physical layer, logical layer, and business layer, and establishes a mapping relationship between the representation layers.

[0071] In practice, hardware configuration and initial operating status information of heterogeneous devices in the target server cluster are collected through standardized interfaces. Based on the collected information, digital representations of the physical, logical, and business layers are generated, and a mapping relationship is established between the three layers to form a digital twin model that is synchronized with the physical server cluster in real time. This model is the foundation for all subsequent simulation analysis, state comparison, and policy verification.

[0072] In one implementation, step S100 specifically includes the following steps:

[0073] S110: Collect hardware configuration information of heterogeneous devices in the target server cluster through a standardized interface. The hardware configuration information includes at least device identifier, performance parameters, and physical location information.

[0074] In practical implementation, standardized interfaces include, but are not limited to, server management protocol interfaces such as IPMI (Intelligent Platform Management Interface), Redfish, and SNMP (Simple Network Management Protocol). Through these interfaces, hardware configuration information is collected from each physical server, switch, storage device, and other heterogeneous device in the target server cluster. Among these, device identifiers are used to uniquely identify each device, such as serial numbers or UUIDs; performance parameters include CPU model and number of cores, memory capacity, disk capacity and type, network card bandwidth, etc.; physical location information includes rack number, rack location, data center room number, etc. This information constitutes the basic data for the physical layer of the digital twin model.

[0075] S120: Based on the hardware configuration information, generate a digital representation of the physical layer, wherein the digital representation of the physical layer includes logical mapping nodes of each of the heterogeneous devices.

[0076] In practice, each heterogeneous device collected is mapped as a logical mapping node in the digital space. Each logical mapping node contains complete hardware configuration information for that device and is stored in the digital twin model as a data object. For example, the logical mapping node of a physical server records information such as the number of CPU cores, memory capacity, disk configuration, number of network cards, and bandwidth. All logical mapping nodes are categorized and tagged according to device type (server, switch, storage, firewall, etc.) and stored in the physical layer node set, forming the physical layer of the digital twin model.

[0077] S130: Identify the physical connection relationships and business call relationships between each of the logical mapping nodes, and construct a resource dependency graph representing the association between resources as a digital representation of the logical layer.

[0078] In practical implementation, the identification of physical connections is achieved through a combination of link-layer discovery protocols and network configuration information: the acquisition engine obtains LLDP or CDP neighbor information of the switch via SNMP, and parses out the physical connections between the switch ports and the server network cards and storage front-end ports; simultaneously, it reads the switch's MAC address table, ARP table, and the server's routing table to complete the port-level topology. Each physical connection is abstracted as a physical connection edge in the resource dependency graph, with attributes including the identifiers of the nodes at both ends of the connection, the port names of the connections, the link bandwidth, the link aggregation relationship, and the current link status. The identification of service call relationships relies on a distributed link tracing and application performance monitoring system: a lightweight proxy is integrated into the service applications of the server cluster or a service mesh (Sidecar) is used to collect trace information for each service call, including the call source, target, protocol, request frequency, and average response time. By aggregating the call data within a certain time window, stable inter-service call dependencies are extracted to generate service call edges. The attributes of the service call edges include the logical mapping nodes of the source and target services, the communication protocol, the QPS baseline, and the maximum tolerable latency. Furthermore, the dependency relationship between the database and the application can be verified by parsing the database connection string and slow query logs. Finally, using logical mapping nodes as vertices and physical connection edges and business call edges as two different types of relationship edges, a directed resource dependency graph with attributes is constructed. This graph can fully express the multi-layered dependency structure from the underlying physical connection to the upper-layer application call, which is the logical layer of the digital twin model.

[0079] S140: Based on the nodes and relationships in the resource dependency graph, generate a visual representation framework for the digital twin model, which serves as the digital representation of the business layer.

[0080] In its implementation, the visualization framework does not simply lay out graph data; instead, it uses resource dependency graphs as the basic data source and customizes perspectives for different operations and maintenance roles. The framework adopts a front-end / back-end separation architecture. The back-end service queries dependencies and dynamic node attributes from the graph database in real time, while the front-end renders based on graphics technologies such as WebGL or Three.js. The system generates three core views by default: a physical topology view, which uses the actual rack layout of a data center as a background, placing logically mapped nodes according to their physical coordinates and using cables to represent physical connections, allowing operations and maintenance personnel to intuitively see the location and connection status of devices on the rack; a logical dependency view, which uses a force-oriented layout to show the call relationships between services, with node size indicating load level and edge thickness indicating call traffic volume. When a service call edge is selected, all physical connection edges it traverses are highlighted, tracing the end-to-end communication path; and a business health view, which uses business systems as aggregation units, displaying the health score of each business system and supporting drill-down to all microservices and their dependent resources involved in that business. When a node or connection is abnormal, the impact is propagated through color changes. The visualization framework also provides a timeline playback function, which can load historical state data to replay the cluster operation status at any time in the past.

[0081] S150: Establish the mapping relationship between the logical mapping node, the resource dependency graph and the visualization representation framework to form an association mapping between the physical layer, the logical layer and the business layer.

[0082] In practice, each logical mapping node is assigned a unique identifier, and the same identifier is used for reference in both the resource dependency graph and the visualization representation framework. When the state of a logical mapping node in the physical layer changes, the corresponding node state in the resource dependency graph in the logical layer can be automatically and synchronously updated through identifier association, and reflected in real time in the visualization representation framework of the business layer. This three-layer association mapping mechanism ensures the consistency between the digital twin model and the physical server cluster.

[0083] S200: Monitor the load indicators of the target server cluster in real time, and determine whether the current time period is an idle or busy period based on the load indicators.

[0084] In practice, real-time load metrics of the target server cluster are continuously collected, and two operating states, idle and busy, are distinguished according to preset judgment rules. Idle state refers to the period when the overall load of the server cluster is low and computing resources are sufficient, which is suitable for offline simulation analysis and knowledge base construction; busy state refers to the period when the server cluster load is high and business pressure is high, which requires prioritizing the response speed of online monitoring and real-time decision-making.

[0085] In one implementation, step S200 specifically includes the following steps:

[0086] S210: Obtain the real-time load metrics of the target server cluster, wherein the real-time load metrics include at least one or more of CPU utilization, memory usage, network throughput, and I / O latency.

[0087] In practice, a monitoring agent deployed on each server collects load metrics such as CPU utilization, memory usage, network throughput, and disk I / O latency at a preset collection frequency (e.g., once every 10 seconds). The load metrics at the cluster level are the weighted average of the corresponding metrics across all server nodes, with the weights set according to the computing power or business importance of each server.

[0088] S220: Compare the real-time load index with a preset idle / busy threshold. When the real-time load index is lower than the idle / busy threshold, determine that the current period is in an idle state; when the real-time load index reaches or exceeds the idle / busy threshold, determine that the current period is in a busy state.

[0089] In practice, the idle / busy threshold can be set with multiple composite conditions. For example, when the cluster's average CPU utilization is below 30% and average memory usage is below 50%, it is considered an idle state; otherwise, it is considered a busy state. The specific value of the threshold can be configured according to the actual business scenario.

[0090] In a more optimized implementation, the idle / busy status determination threshold is obtained by extracting historical load data from the status database, analyzing the historical load data to identify the periodic patterns of the target server cluster's load (such as periodic changes in daytime peaks and nighttime troughs), and dynamically adjusting the idle / busy status determination threshold based on these patterns. For example, the idle status determination threshold can be appropriately increased for nighttime periods and appropriately decreased for daytime peak business periods to improve the accuracy and adaptability of idle / busy status determination.

[0091] S300: When the system is determined to be in an idle state, the offline simulation and knowledge construction process is executed.

[0092] In practice, offline simulation and knowledge building are performed during off-peak hours. This process utilizes periods of low server cluster load to simulate system behavior under various operating conditions by injecting pre-set load and fault data into the digital twin model. Based on the simulation results, a state database is built and improved, and dynamic judgment criteria and candidate operation and maintenance strategies are generated and optimized. This process provides a decision-making basis for rapid response during peak hours.

[0093] In one implementation, step S300 specifically includes the following steps:

[0094] S310: Input a preset load dataset and fault event set into the digital twin model to simulate the evolution of resource status and business status, and generate simulation analysis results.

[0095] In specific implementation, step S310 further includes:

[0096] S311: Obtain the preset load strategy and fault strategy, generate the corresponding load dataset and fault event set, and inject the load dataset and fault event set into the digital twin model.

[0097] The load strategy includes a service load simulation strategy and a deployment and launch simulation strategy, and the fault strategy includes a hardware fault injection strategy and a network fault injection strategy.

[0098] In practical implementation, the business load simulation strategy is used to simulate the pressure on the server cluster caused by real business requests, such as simulating the instantaneous high-concurrency traffic of a "Double Eleven" flash sale scenario or simulating the periodic fluctuations of daily business volume. The deployment and launch simulation strategy is used to simulate resource changes during the launch of new services or the migration of existing services, such as simulating the elastic scaling of containerized applications or simulating virtual machine migration. The hardware fault injection strategy is used to simulate the fault behavior of server hardware components, such as CPU throttling, uncorrectable memory errors, and bad sectors on the disk. The network fault injection strategy is used to simulate network-level anomalies, such as increased network latency, increased packet loss rate, and network partitioning.

[0099] S312: Drive the digital twin model to run a simulation based on the injected data, and record the resource status change sequence and business call status change sequence of each node during the simulation.

[0100] The resource status change sequence includes at least resource consumption trends, resource conflict markers, and conflict node information, and the service call status change sequence includes at least service call path changes and cascading impact range.

[0101] In practical implementation, after injecting load and fault data, the digital twin model simulates the state changes of each node step by step over time. The resource state change sequence records the changes in resource indicators such as CPU utilization, memory usage, disk I / O, and network bandwidth for each node at each simulation time step, and marks the nodes and time points where resource preemption or bottlenecks occur. The service call state change sequence records the path changes of service requests in the service call chain. When a node fails, it records the cascading effects caused by the failure, such as call chain rerouting and downstream service timeouts.

[0102] S313: Generate the simulation analysis results based on the resource status change sequence and the service call status change sequence. The simulation analysis results include at least one of resource bottleneck analysis, capacity prediction, fault impact range assessment, or deployment scheme verification.

[0103] In practice, resource bottleneck analysis identifies the resource nodes and resource types that reach their performance limits first under specific load conditions; capacity prediction extrapolates the cluster's expansion time and scale requirements based on simulation data under business growth trends; fault impact assessment quantifies the scope of business affected and the degree of service degradation after a specific node or component fails; and deployment scheme verification rehearses the effectiveness of new deployment schemes in a simulation environment and evaluates their resource consumption and performance.

[0104] S320: Construct and update the state database based on the simulation analysis results. The state database stores the normal state mode, abnormal state characteristics and performance baseline under different operating conditions.

[0105] In practical implementation, feature extraction is performed on the simulation analysis results to obtain the normal state pattern, abnormal state characteristics, and performance baseline under the simulated operating conditions. The normal state pattern refers to the normal fluctuation range and timing pattern of various indicators of the system under specific load conditions; abnormal state characteristics refer to the abnormal performance and combination characteristics of indicators corresponding to various fault types; and the performance baseline refers to the benchmark values ​​of the system's performance indicators under standard test conditions. The extracted patterns, characteristics, and baselines are then integrated and stored with existing historical data in the state database to expand the operating condition range covered by the state database.

[0106] S330: Based on the simulation analysis results and historical operation data, generate and optimize dynamic judgment criteria and candidate operation and maintenance strategy set.

[0107] In practical implementation, dynamic judgment criteria include dynamic alarm thresholds and anomaly detection models. Based on the boundary characteristics of normal and abnormal states in the simulation analysis results, the alarm thresholds are adjusted so that alarm triggering can promptly reflect anomalies without generating too many false alarms. The candidate operation and maintenance strategy set is a predefined collection of executable operation and maintenance operations. The repair effects of each strategy under different fault scenarios are evaluated based on the simulation analysis results, and the strategies are ranked, filtered, and optimized.

[0108] S400: When the system is determined to be in a busy state, it performs online monitoring and real-time decision-making processes.

[0109] In practice, during peak business periods, the process directly utilizes the status database and dynamic judgment criteria built during off-peak hours to detect anomalies by rapidly comparing real-time data with historical patterns. After simulating and verifying candidate strategies in the digital twin model, the strategies are then deployed for execution, ensuring the timeliness and accuracy of the operation and maintenance response.

[0110] In one implementation, step S400 specifically includes the following steps:

[0111] S410: Obtain the real-time operating status data of the target server cluster, compare the real-time operating status data with the normal status mode in the status database, and generate an alarm signal based on the current dynamic judgment criteria.

[0112] In specific implementation, step S410 further includes:

[0113] S411: Synchronize the collected real-time operating status data to the digital twin model to drive the digital twin model to perform real-time monitoring of multi-dimensional performance indicators of the target server cluster.

[0114] In practice, real-time operational status data is collected by a monitoring agent deployed on the server and synchronously transmitted to the corresponding logical mapping node in the digital twin model to update the real-time status of the node, so that the digital twin model and the physical server cluster maintain real-time synchronization.

[0115] S412: Extract historical normal state patterns, indicator trend characteristics and periodic characteristics from the state database, and combine them with the current real-time operating data to generate dynamic alarm thresholds and anomaly detection models as the dynamic judgment criteria.

[0116] In practical implementation, anomaly detection models can be constructed using statistical or machine learning methods. For example, normal fluctuation ranges for various performance indicators can be established based on historical data, and an alarm can be triggered when real-time data exceeds the range; or an autoencoder model can be trained to use reconstruction error as a measure of the degree of anomaly.

[0117] S413: Input the real-time monitoring data of the multi-dimensional performance indicators into the dynamic judgment criteria to determine whether the alarm triggering conditions are met, and generate an alarm signal carrying the alarm level and the scope of influence when the conditions are met.

[0118] In practice, alarm levels are categorized based on the severity of the anomaly, such as alert, warning, severe, and urgent. The scope of impact is assessed based on the resource dependency graph and the location of the faulty node, such as affecting a single server, a specific business module, or the entire cluster.

[0119] In a more optimized implementation, the specific method for comparing real-time operational status data with normal status patterns in the status database is as follows: Multi-dimensional performance indicators are extracted from the real-time operational status data to form a real-time status feature vector; a normal status pattern matching the current operating condition is retrieved from the status database to form a baseline feature vector; the deviation between the real-time status feature vector and the baseline feature vector is calculated, and further anomaly identification is triggered when the deviation exceeds a preset range. Specifically, the normal status pattern matching the current operating condition is determined by performing a multi-dimensional search in the status database based on the current time window, service type, and load level of the real-time operational status data, retrieving the historical normal status pattern with the highest similarity to the current time window, service type, and load level.

[0120] S420: In response to the alarm signal, a target operation and maintenance strategy is matched from the candidate operation and maintenance strategy set, and the target operation and maintenance strategy is simulated and verified in the digital twin model.

[0121] In specific implementation, step S420 further includes:

[0122] S421: Extract key feature information from the alarm signal. The key feature information includes at least the fault type identifier, the range of affected resources, and the degree of performance degradation.

[0123] In practice, the fault type identifier is used to indicate the type of fault, such as CPU overload, memory leak, network congestion, disk failure, etc.; the affected resource range indicates the list of server nodes or services affected by the fault; and the performance degradation level quantifies the degree of impact of the fault on business performance, such as the percentage increase in response time or the percentage decrease in throughput.

[0124] S422: Match the key feature information with a preset candidate operation and maintenance strategy library to obtain at least one candidate operation and maintenance strategy, and determine the target operation and maintenance strategy from it according to the preset matching degree sorting rules.

[0125] In practice, the candidate operation and maintenance strategy library stores applicable scenario tags for each strategy, including applicable fault types, resource scope, and severity. The matching process retrieves strategies that match the applicable scenario tags based on key feature information and sorts them from highest to lowest matching degree. The matching degree can be calculated comprehensively based on multiple factors such as historical execution success rate, strategy execution time, and the degree of impact of the strategy on business. The strategy with the highest matching degree is selected as the target operation and maintenance strategy.

[0126] S423: Simulate the execution of the target operation and maintenance strategy in the digital twin model and obtain the simulation execution results, which include changes in resource status and service recovery status after execution.

[0127] In practical implementation, a sandbox environment is created within the digital twin model, using the current real-time state of the system as the initial state to simulate the execution of the target operation and maintenance strategy. For example, if the target operation and maintenance strategy is "to migrate service A from node X to node Y," the migration operation is simulated in the sandbox environment, and the changes in resource utilization of each node and the recovery of service A's response time are observed after the migration.

[0128] S424: Compare the simulation execution result with the preset verification conditions, and determine whether the target operation and maintenance strategy passes the verification based on the comparison result.

[0129] In practice, the preset verification conditions include: whether the resource utilization rate of the target node is within a safe range after execution, whether the performance indicators of the affected services have recovered to normal levels, and whether new resource conflicts or cascading failures have been triggered. Only when the simulation execution result meets all verification conditions is the target operation and maintenance strategy deemed to have passed verification.

[0130] S430: The verified target operation and maintenance strategy is sent to the target server cluster for execution, and the execution result information is obtained.

[0131] In specific implementation, step S430 further includes:

[0132] S431: Convert the verified target operation and maintenance strategy into an operation instruction set recognizable by the target server cluster, and send the operation instruction set to the target server cluster through a standardized interface.

[0133] In practice, the target operation and maintenance strategy is stored in an abstract descriptive form (such as "migration service A"), which needs to be converted into specific operation instructions that the target server cluster management platform can recognize (such as calling the Kubernetes API to perform Pod migration, calling the virtualization management platform API to perform online virtual machine migration, etc.). After the instructions are converted, they are issued and executed through the corresponding standardized API interface.

[0134] S432: During the execution process, continuously acquire real-time running status data of the target server cluster and compare it in real time with the simulated execution results in the digital twin model to monitor whether the execution process deviates from expectations.

[0135] In practice, an execution monitoring window is set up to continuously collect cluster operational status data for a period of time after the command is issued. The change curve of the actual operational status data is compared in real time with the predicted change curve of the simulated execution result in the sandbox simulation to calculate the degree of deviation. If the deviation exceeds a preset threshold, it is determined that the execution process has deviated from expectations, triggering manual intervention or a rollback operation.

[0136] S433: After the operation instruction set is executed, the actual running status data of the target server cluster is obtained as the execution result information. The execution result information includes at least resource status changes, service recovery status and execution time.

[0137] In practice, the execution results are recorded for subsequent updates to the status database and optimization of the candidate operation and maintenance strategy set.

[0138] S500: Store the full lifecycle status information, log information, alarm information, and operation and maintenance strategy execution information obtained during the idle and busy periods into the status database.

[0139] In practice, the state database serves not only as a storage medium for offline simulation results during idle periods but also as a storage medium for all actual operational data generated during online maintenance during busy periods. By continuously accumulating lifecycle data into the state database, a constantly growing and self-optimizing knowledge base is formed.

[0140] In one implementation, step S500 specifically includes the following steps:

[0141] S510: Obtain multi-source heterogeneous operation and maintenance data generated during the idle and busy periods. The multi-source heterogeneous operation and maintenance data includes at least configuration status information, operation log information, historical alarm information, and operation and maintenance strategy execution information.

[0142] In practice, configuration status information records snapshots of the server cluster's hardware and software configurations at various points in time; operation log information records the operation logs of the system and applications; historical alarm information records the trigger time, alarm content, alarm level, and processing result of each alarm; and operation and maintenance policy execution information records the time, policy content, execution process, and execution effect of each policy execution.

[0143] S520: Add timestamps and event association tags to the multi-source heterogeneous operation and maintenance data, establish temporal and causal relationships between the operation and maintenance data, and form a structured operation and maintenance event chain.

[0144] In practice, a time-series index of the data is established based on timestamps. Event association tags are used to mark the causal relationships between different data. For example, alarm information is associated with the anomaly log that triggered the alarm, and operation and maintenance policy execution information is associated with the alarm information responded to by the policy. Through the above associations, a complete operation and maintenance event chain is formed from "anomaly occurrence → alarm triggering → policy execution → effect feedback".

[0145] S530: Based on the operation and maintenance event chain, extract operation and maintenance scenario features, fault root cause features and strategy effect evaluation indicators, generate an operation and maintenance knowledge structure containing feature tags and retrieval indexes, and store it in the status database.

[0146] In practical implementation, the operation and maintenance scenario features describe the business background and system status at the time of the event; the root cause features extract the root cause of the failure and the failure propagation path; and the strategy effectiveness evaluation indicators quantify the effectiveness of strategy execution, such as failure recovery time and the degree of impact on business. These features and indicators are stored in the form of structured knowledge entries, supporting multi-dimensional retrieval by scenario, failure type, strategy effectiveness, and other dimensions.

[0147] S600: Using the information in the state database, iteratively update the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set.

[0148] In practice, as the actual operation and maintenance data in the status database continues to accumulate, this data is used to continuously iterate and optimize the core components of the system, forming a closed-loop self-learning mechanism.

[0149] In one implementation, step S600 specifically includes the following steps:

[0150] S610: Perform a deviation analysis between the actual operating data obtained during the idle and busy periods and the simulation prediction data of the digital twin model.

[0151] In practice, actual operational data is periodically (e.g., weekly or monthly) extracted from the status database and compared with the simulation prediction data of the digital twin model within the corresponding time window to calculate the prediction deviation. For example, for a certain fault event, the difference between the fault impact range predicted by the digital twin model and the actual fault impact range is compared.

[0152] S620: Based on the deviation analysis results, adjust the model parameters and mapping relationships in the digital twin model to keep the digital twin model synchronized with the actual operating state of the target server cluster.

[0153] In practice, if deviation analysis reveals discrepancies between the model's resource consumption parameters and actual data, the model parameters are adjusted based on the actual data. If the mapping relationship fails to accurately reflect new business dependencies, the resource dependency graph is updated. Through continuous deviation analysis and parameter adjustments, the digital twin model gradually improves the simulation accuracy of the physical server cluster.

[0154] S630: Based on the newly added operation and maintenance event chain data in the status database, update the alarm threshold and anomaly detection model parameters in the dynamic judgment criteria.

[0155] In practice, newly added normal operation data and abnormal event data are used as training samples to retrain the anomaly detection model and update the alarm thresholds regularly, so that the alarm judgment criteria can adapt to changes in the system's operating status.

[0156] S640: Based on the newly added strategy execution effect data in the status database, update the matching degree ranking weight and applicable conditions of each strategy in the candidate operation and maintenance strategy set.

[0157] In practice, if a strategy performs well multiple times in actual execution, its matching weight is increased; if a strategy has had adverse consequences in actual execution, its weight is reduced or usage restrictions are added to ensure the accuracy and security of strategy matching.

[0158] The data processing method of this invention, through steps S100 to S600, realizes a complete closed-loop process from digital twin model construction to intelligent operation and maintenance throughout the entire lifecycle. The method utilizes load indicators to distinguish between idle and busy times. Offline simulation and knowledge base construction are performed during idle times, while online monitoring and real-time decision-making are conducted during busy times. The state database enables continuous knowledge accumulation and iterative optimization of core system components. This method effectively utilizes the idle computing resources of the server cluster for simulation and deduction, while ensuring the timeliness and accuracy of operation and maintenance responses during busy times. It has significant advantages such as high automation, accurate operation and maintenance decisions, and strong adaptability.

[0159] Example 2

[0160] A server lifecycle intelligent operation and maintenance system based on digital twins includes:

[0161] The model building module is used to obtain the physical configuration information and running status information of the target server cluster, construct a digital twin model corresponding to the target server cluster, the digital twin model includes digital representation of at least one of the physical layer, logical layer and business layer, and establishes the mapping relationship between the representation layers;

[0162] The load status judgment module is used to detect the load index of the target server cluster and determine whether the current running status is an idle state or a busy state based on the load index.

[0163] The idle time processing module is used to perform offline simulation and knowledge base construction when the state is determined to be idle: inputting a preset load dataset and fault event set into the digital twin model, simulating the evolution of resource status and business status, and generating simulation analysis results; constructing and updating a status database based on the simulation analysis results, the status database storing normal status patterns, abnormal status characteristics and performance baselines under different operating conditions; and generating and optimizing dynamic judgment criteria and candidate operation and maintenance strategy sets according to the simulation analysis results and historical operation data.

[0164] The busy-hour processing module is used to perform online monitoring and real-time decision-making processes when a busy-hour state is determined: acquiring real-time operating status data of the target server cluster, comparing the real-time operating status data with the normal state mode in the status database, and generating an alarm signal based on the current dynamic judgment criteria; responding to the alarm signal, matching a target operation and maintenance strategy from the candidate operation and maintenance strategy set, and simulating and verifying the target operation and maintenance strategy in the digital twin model; and sending the verified target operation and maintenance strategy to the target server cluster for execution and obtaining execution result information.

[0165] The information storage module is used to store the full lifecycle status information, log information, alarm information and operation and maintenance strategy execution information obtained in the idle and busy states into the status database;

[0166] The iterative update module is used to iteratively update the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set using information from the state database.

[0167] Example 3

[0168] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0169] Figure 2A schematic block diagram of an example electronic device 100 that can be used to implement embodiments of the present invention is shown. Electronic device 100 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 100 may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their links and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0170] like Figure 2 As shown, the electronic device 100 includes a computing unit 101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 102 or a computer program loaded from a storage unit 108 into a random access memory (RAM) 103. The RAM 103 may also store various programs and data required for the operation of the device 100. The computing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also linked to the bus 104.

[0171] Multiple components in electronic device 100 are linked to I / O interface 105, including: input unit 106, such as keyboard, mouse, etc.; output unit 107, such as various types of displays, speakers, etc.; storage unit 108, such as disk, optical disk, etc.; and communication unit 109, such as network card, modem, wireless transceiver, etc. Communication unit 109 allows electronic device 100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0172] The computing unit 101 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 101 performs the various methods and processes described above, such as a digital twin-based intelligent operation and maintenance method for the entire lifecycle of a server. For example, in some embodiments, a digital twin-based intelligent operation and maintenance method for the entire lifecycle of a server can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 108. In some embodiments, part or all of the computer program can be loaded and / or installed on device 100 via ROM 102 and / or communication unit 109. When the computer program is loaded into RAM 103 and executed by the computing unit 101, one or more steps of the digital twin-based intelligent operation and maintenance method for the entire lifecycle of a server can be performed. Alternatively, in other embodiments, the computing unit 101 may be configured, by any other suitable means (e.g., by means of firmware), to perform a digital twin-based intelligent operation and maintenance method for the entire lifecycle of a server.

[0173] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0174] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0175] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical links based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0176] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0177] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0178] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0179] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0180] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A server full lifecycle intelligent operation and maintenance method based on digital twins, characterized in that, include: Obtain the physical configuration information and running status information of the target server cluster, construct a digital twin model corresponding to the target server cluster, the digital twin model includes digital representations of at least one of the physical layer, logical layer and business layer, and establish mapping relationships between the representation layers; Detect the load index of the target server cluster, and determine whether the current operating state is an idle state or a busy state based on the load index; When the system is determined to be in an idle state, an offline simulation and knowledge base construction process is executed: a preset load dataset and fault event set are input into the digital twin model to simulate the evolution of resource status and business status and generate simulation analysis results. Based on the simulation analysis results, a state database is constructed and updated. The state database stores normal state modes, abnormal state characteristics and performance baselines under different operating conditions. Furthermore, based on the simulation analysis results and historical operation data, dynamic judgment criteria and candidate operation and maintenance strategy sets are generated and optimized; When a busy state is detected, an online monitoring and real-time decision-making process is executed: real-time operating status data of the target server cluster is acquired, the real-time operating status data is compared with the normal state mode in the status database, and an alarm signal is generated according to the current dynamic judgment criteria; in response to the alarm signal, a target operation and maintenance strategy is matched from the candidate operation and maintenance strategy set, and the target operation and maintenance strategy is simulated and verified in the digital twin model; the verified target operation and maintenance strategy is sent to the target server cluster for execution, and the execution result information is obtained; The full lifecycle status information, log information, alarm information, and operation and maintenance strategy execution information obtained during the idle and busy periods are stored in the status database. Using the information in the state database, the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set are iteratively updated.

2. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, The process involves acquiring the physical configuration and operational status information of the target server cluster, constructing a digital twin model corresponding to the target server cluster, wherein the digital twin model includes digital representations of at least one of the physical, logical, and business layers, and establishing mapping relationships between the representation layers. Specifically, this includes: Hardware configuration information of heterogeneous devices in the target server cluster is collected through a standardized interface. The hardware configuration information includes at least device identifier, performance parameters and physical location information. Based on the hardware configuration information, a digital representation of the physical layer is generated, and the digital representation of the physical layer includes logical mapping nodes for each of the heterogeneous devices. Identify the physical connection relationships and business call relationships between each of the logical mapping nodes, and construct a resource dependency graph representing the association between resources as a digital representation of the logical layer; Based on the nodes and relationships in the resource dependency graph, a visualization representation framework for the digital twin model is generated, which serves as the digital representation of the business layer. Establish the mapping relationship between the logical mapping node, the resource dependency graph and the visualization representation framework to form an association mapping between the physical layer, the logical layer and the business layer.

3. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, The process of inputting a preset load dataset and fault event set into the digital twin model to simulate the evolution of resource and business states in the digital twin model and generate simulation analysis results specifically includes: Obtain preset load policies and fault policies, generate corresponding load datasets and fault event sets, and inject the load datasets and fault event sets into the digital twin model. The load policies include business load simulation policies and deployment and online simulation policies, and the fault policies include hardware fault injection policies and network fault injection policies. The digital twin model is driven to run a simulation based on the injected data, and the resource status change sequence and service call status change sequence of each node are recorded during the simulation. The resource status change sequence includes at least resource consumption trend, resource conflict marker and conflict node information, and the service call status change sequence includes at least service call path change and cascading impact range. The simulation analysis results are generated based on the resource status change sequence and the service call status change sequence. The simulation analysis results include at least one of resource bottleneck analysis, capacity prediction, fault impact range assessment, or deployment scheme verification.

4. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, The step of comparing the real-time operating status data with the normal status patterns in the status database and generating an alarm signal based on the current dynamic judgment criteria specifically includes: The collected real-time operating status data is synchronized to the digital twin model to drive the digital twin model to perform real-time monitoring of multi-dimensional performance indicators of the target server cluster. Historical normal state patterns, indicator trend characteristics, and periodic characteristics are extracted from the state database and combined with the current real-time operating data to generate dynamic alarm thresholds and anomaly detection models, which serve as the dynamic judgment criteria. The real-time monitoring data of the multi-dimensional performance indicators are input into the dynamic judgment criteria to determine whether the alarm triggering conditions are met, and an alarm signal carrying the alarm level and impact range is generated when the conditions are met.

5. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, In response to the alarm signal, the process of matching a target operation and maintenance strategy from the candidate operation and maintenance strategy set and performing simulation verification of the target operation and maintenance strategy in the digital twin model specifically includes: Extract key feature information from the alarm signal, including at least the fault type identifier, the range of affected resources, and the degree of performance degradation; The key feature information is matched with a preset candidate operation and maintenance strategy library to obtain at least one candidate operation and maintenance strategy, and the target operation and maintenance strategy is determined from it according to a preset matching degree sorting rule. The target operation and maintenance strategy is simulated and executed in the digital twin model to obtain simulation execution results, which include changes in resource status and service recovery status after execution. The simulation execution results are compared with preset verification conditions, and the target operation and maintenance strategy is determined to pass the verification based on the comparison results.

6. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, The step of distributing the verified target operation and maintenance strategy to the target server cluster for execution and obtaining execution result information specifically includes: The verified target operation and maintenance strategy is converted into an operation instruction set that can be recognized by the target server cluster, and the operation instruction set is sent to the target server cluster through a standardized interface; During the execution process, the real-time running status data of the target server cluster is continuously acquired and compared in real time with the simulated execution results in the digital twin model to monitor whether the execution process deviates from expectations. After the operation instruction set is executed, the actual running status data of the target server cluster is obtained as the execution result information. The execution result information includes at least resource status changes, service recovery status, and execution time.

7. The intelligent operation and maintenance method for the entire lifecycle of a server based on digital twins according to claim 1, characterized in that, The step of storing the full lifecycle status information, log information, alarm information, and operation and maintenance strategy execution information obtained during idle and busy periods into the status database specifically includes: Acquire multi-source heterogeneous operation and maintenance data generated during idle and busy periods. The multi-source heterogeneous operation and maintenance data includes at least configuration status information, operation log information, historical alarm information, and operation and maintenance strategy execution information. Add timestamps and event association tags to the multi-source heterogeneous operation and maintenance data to establish temporal and causal relationships between the operation and maintenance data, forming a structured operation and maintenance event chain; Based on the operation and maintenance event chain, operation and maintenance scenario features, fault root cause features, and strategy effectiveness evaluation indicators are extracted to generate an operation and maintenance knowledge structure containing feature tags and retrieval indexes, which is then stored in the status database.

8. A server full lifecycle intelligent operation and maintenance system based on digital twins, characterized in that, include: The model building module is used to obtain the physical configuration information and running status information of the target server cluster, construct a digital twin model corresponding to the target server cluster, the digital twin model includes digital representation of at least one of the physical layer, logical layer and business layer, and establishes the mapping relationship between the representation layers; The load status judgment module is used to detect the load index of the target server cluster and determine whether the current running status is an idle state or a busy state based on the load index. The idle time processing module is used to perform offline simulation and knowledge base construction process when the idle time state is determined: input the preset load dataset and fault event set into the digital twin model, simulate the evolution process of resource state and business state, and generate simulation analysis results; Based on the simulation analysis results, a state database is constructed and updated. The state database stores normal state modes, abnormal state characteristics and performance baselines under different operating conditions. Furthermore, based on the simulation analysis results and historical operation data, dynamic judgment criteria and candidate operation and maintenance strategy sets are generated and optimized; The busy-hour processing module is used to perform online monitoring and real-time decision-making processes when a busy-hour state is determined: acquiring real-time operating status data of the target server cluster, comparing the real-time operating status data with the normal state mode in the status database, and generating an alarm signal based on the current dynamic judgment criteria; in response to the alarm signal, matching the target operation and maintenance strategy from the candidate operation and maintenance strategy set, and simulating and verifying the target operation and maintenance strategy in the digital twin model; The verified target operation and maintenance strategy is distributed to the target server cluster for execution, and the execution result information is obtained. The information storage module is used to store the full lifecycle status information, log information, alarm information and operation and maintenance strategy execution information obtained in the idle and busy states into the status database; The iterative update module is used to iteratively update the digital twin model, the dynamic judgment criteria, and the candidate operation and maintenance strategy set using information from the state database.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the server full lifecycle intelligent operation and maintenance method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that enables a processor to execute the intelligent operation and maintenance method for the entire lifecycle of a server as described in any one of claims 1-7.

Images