Service fault recovery method and system based on disaster recovery strategy optimization

Abstract

The invention provides a service fault recovery method and system based on disaster recovery strategy optimization, and the method comprises the steps: obtaining a network topological graph and an enterprise digital twin model based on an enterprise IT architecture; obtaining an initial disaster recovery strategy based on the network topological graph; inputting the initial disaster recovery strategy to the enterprise digital twin model, optimizing the initial disaster recovery strategy until a preset convergence requirement is met, obtaining a target disaster recovery strategy, and storing the target disaster recovery strategy to a preset cloud strategy library; and when a service fault occurs, automatically switching a target disaster recovery strategy based on the preset cloud strategy library so as to recover the service fault. According to the invention, the problem of poor service fault recovery reliability caused by the fact that the disaster recovery scheme cannot be adaptively adjusted according to different service fault scenes in the prior art is solved. According to the invention, self-adaptive adjustment of the disaster recovery strategy is realized, and service fault recovery reliability is improved.

Description

Technical Field

[0001] The present invention relates to the field of disaster recovery and fault recovery, and in particular to a business fault recovery method and system based on disaster recovery strategy optimization. Background Art

[0002] Cloud computing, big data, and artificial intelligence are emerging technologies driving the digital transformation of enterprise IT architectures. Modern infrastructure is dynamic, heterogeneous, and distributed, significantly increasing the probability that systems will face compound risks such as hardware failures, APT attacks, regional natural disasters, and distributed software defects. Therefore, ensuring that enterprise IT architectures can quickly respond to and recover from business failures has become a key issue.

[0003] Currently, existing technologies generally use regular backups combined with manual drills to respond to and recover from business failures. Traditional disaster recovery drills rely on manual labor and the business recovery time is long. Therefore, the existing static plan configuration mechanism is difficult to adaptively adjust the disaster recovery plan according to different business failure scenarios, resulting in poor reliability of business failure recovery. Summary of the Invention

[0004] In order to solve the above problems, the present invention proposes a service failure recovery method and system based on disaster recovery strategy optimization, which realizes adaptive adjustment of disaster recovery strategy and improves the reliability of service failure recovery.

[0005] To achieve the above-mentioned objectives, an embodiment of the present invention provides a business fault recovery method based on disaster recovery strategy optimization, including: obtaining a network topology map and an enterprise digital twin model based on the enterprise IT architecture; obtaining an initial disaster recovery strategy based on the network topology map; inputting the initial disaster recovery strategy into the enterprise digital twin model, optimizing the initial disaster recovery strategy until the preset convergence requirements are met, obtaining a target disaster recovery strategy and storing the target disaster recovery strategy in a preset cloud policy library; when a business failure occurs, automatically switching the target disaster recovery strategy based on the preset cloud policy library to recover the business failure.

[0006] An embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, constructs a linkage mechanism between the enterprise digital twin model and the network topology map, inputs the initial disaster recovery strategy into the enterprise digital twin model for simulation iterative optimization, and utilizes the real-time mapping and prediction capabilities of the enterprise digital twin model to make the target disaster recovery strategy deeply adapt to the risk characteristics of different business scenarios. At the same time, the target disaster recovery strategy is also stored in a preset cloud policy library. When a fault occurs, intelligent matching and automatic switching of multi-dimensional disaster recovery plans are realized to recover business faults, breaking through the limitations of traditional static plans. Therefore, the reliability of business fault recovery is improved through continuous optimization of virtual-reality interaction and adaptive adjustment of cloud strategies.

[0007] Furthermore, based on the enterprise IT architecture, a network topology map and an enterprise digital twin model are obtained, including: setting device nodes and connecting device nodes based on the enterprise IT architecture to obtain a network topology map; constructing a resource load change model based on the first resource load parameters of each device node; constructing a network communication model based on the second resource load parameters of each device node; and generating an enterprise digital twin model corresponding to the network topology map based on the resource load change model and the network communication model.

[0008] In the above scheme, device nodes are first set up for the enterprise IT architecture and a network topology diagram is constructed. Then, the resource load change model established by the first resource load parameter and the network communication model generated by the second resource load parameter are combined to form an enterprise digital twin model with real-time mapping and prediction capabilities. The initial disaster recovery strategy is dynamically simulated and iteratively optimized through the enterprise digital twin model, so that the strategy can accurately match the resource load status and network communication characteristics of different device nodes, thereby adaptively adjusting the disaster recovery plan under business failure scenarios to recover business failures and improve the reliability of business failure recovery.

[0009] Furthermore, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, which also includes: obtaining real-time device node data based on the network topology diagram; inputting the real-time device node data into the enterprise digital twin model to conduct disaster scenario simulation drills to obtain the fault propagation situation; executing the preset disaster recovery strategy based on the fault propagation situation and calculating the simulation error; based on the simulation error, adjusting the model parameters of the enterprise digital twin model to optimize the enterprise digital twin model.

[0010] In the above scheme, the device node data in the network topology is collected in real time and input into the enterprise digital twin model, dynamically simulating the fault propagation path and impact range in the disaster scenario. Then, based on the error analysis between the simulation results of the preset disaster recovery strategy and the actual fault propagation situation, the resource load and communication parameters of the digital twin model are reversely iterated and adjusted to make the model continuously approach the real business environment. Thus, through the closed-loop feedback mechanism of disaster recovery drills and parameter optimization, the prediction accuracy of the digital twin model for different disaster scenarios is enhanced, thereby improving the adaptability of the disaster recovery strategy to dynamic fault characteristics, realizing policy adaptive optimization driven by real-time data, and significantly improving the reliability of business fault recovery.

[0011] Furthermore, based on the network topology map, an initial disaster recovery strategy is obtained, including: real-time monitoring of the network topology map, obtaining real-time disaster recovery data and acquiring characteristic information corresponding to the real-time disaster recovery data; constructing a state space based on the first characteristic information; constructing an action space based on the second characteristic information; based on the state space, calculating the initial probability of selecting a random action in the action space under any state; and obtaining an initial disaster recovery strategy based on the initial probability.

[0012] In the above scheme, the disaster recovery data in the network topology diagram is monitored in real time, the first characteristic information is extracted to construct a state space to characterize the real-time operating status of the system, and the action space is generated according to the second characteristic information to define a set of executable disaster recovery operations. By calculating the initial probability distribution of random actions in the action space under each state in the state space, an initial disaster recovery strategy that dynamically adapts to the current system state is formed, so that the strategy generation process is closely related to the real-time environmental characteristics to achieve adaptive adjustment of the disaster recovery plan and improve the reliability of business fault recovery.

[0013] Furthermore, the initial disaster recovery strategy is input into the enterprise digital twin model, and the initial disaster recovery strategy is optimized until the preset convergence requirements are met, and the target disaster recovery strategy is obtained and stored in the preset cloud strategy library, including: setting a reward function based on the third characteristic information; inputting the initial disaster recovery strategy into the enterprise digital twin model to perform a simulation drill, and calculating the instant reward for selecting the corresponding action under each drill state through the reward function to obtain a first instant reward; adjusting the initial disaster recovery strategy based on the preset random disturbance, and recalculating the instant reward for selecting the corresponding action under each drill state to obtain a second instant reward; based on the first instant reward and the second instant reward, optimizing the initial disaster recovery strategy until the preset convergence requirements are met, obtaining the target disaster recovery strategy, and storing the target disaster recovery strategy in the preset cloud strategy library.

[0014] In the above scheme, a reward function is first constructed using the third characteristic information. When simulating the initial disaster recovery strategy in the enterprise digital twin model, the reward function is used to quantify the first instant rewards of different actions in the drill state, and the second instant reward is generated through a random perturbation strategy. Then, the initial disaster recovery strategy is adjusted according to the first instant reward and the second instant reward. The reinforcement learning mechanism is used to drive the iterative optimization of the disaster recovery strategy in the virtual environment until convergence, so that the initial disaster recovery strategy is deeply adapted to the complex fault scenario characteristics simulated by the digital twin model. Finally, the target disaster recovery strategy is generated and stored in the preset cloud policy library, forming an adaptive strategy generation closed loop based on real-time drill feedback, breaking through the limitation of the static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business failure scenarios, and improving the reliability of business fault recovery.

[0015] Furthermore, based on the first instant reward and the second instant reward, the initial disaster recovery strategy is optimized until the preset convergence requirements are met, the target disaster recovery strategy is obtained, and the target disaster recovery strategy is stored in the preset cloud strategy library, including: based on the preset reinforcement learning algorithm and the first instant reward, a first action value function is obtained; based on the preset reinforcement learning algorithm and the second instant reward, a second action value function is obtained; based on the second action value function, the initial disaster recovery strategy is updated, and the change value between the first action value function and the second action value function is calculated; if the change value between the first action value function and the second action value function meets the preset convergence requirements, the updated initial disaster recovery strategy is used as the target disaster recovery strategy, and the target disaster recovery strategy is stored in the preset cloud strategy library.

[0016] In the above scheme, a preset reinforcement learning algorithm is used to construct the first action value function and the second action value function based on the first immediate reward and the second immediate reward respectively. The optimization direction of the disaster recovery strategy is dynamically evaluated by comparing the change values ​​of the two, and the initial disaster recovery strategy is iteratively updated in the digital twin environment. When the change values ​​of the first action value function and the second action value function meet the convergence conditions, it indicates that the decision-making benefits of the strategy under the simulated fault scenario tend to be stable. At this time, the updated initial disaster recovery strategy is used as the target disaster recovery strategy and stored in the preset cloud policy library, forming an adaptive strategy generation closed loop based on real-time exercise feedback, breaking through the limitation of the static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business failure scenarios, and improving the reliability of business fault recovery.

[0017] Furthermore, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, which also includes: obtaining actual drill results based on the target disaster recovery strategy; inputting the target disaster recovery strategy into the enterprise digital twin model to perform a simulation drill to obtain simulation results; obtaining a drill error based on the actual drill results and the simulation results; if the drill error does not meet the preset threshold requirements, optimizing the target disaster recovery strategy.

[0018] In the above scheme, the target disaster recovery strategy is input into the digital twin model to perform simulation exercises to generate simulation results, and the error is compared with the actual exercise results. When the error exceeds the preset threshold, the strategy iterative optimization is triggered. Then, by judging whether the exercise error between the actual exercise results and the simulation results meets the threshold requirements, the deviation between the digital twin model prediction and the actual fault recovery process is continuously eliminated, so that the target disaster recovery strategy dynamically approaches the optimal solution of the real business environment in the feedback loop of virtual and real integration, thereby enhancing the adaptability of the target disaster recovery strategy to different fault scenarios and improving the reliability of business fault recovery.

[0019] Furthermore, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, including: obtaining an operating data set based on a network topology diagram; obtaining a long-term trend component based on a preset local regression algorithm and the operating data set; obtaining a periodic component based on a preset fitting algorithm and the operating data set; obtaining a residual term through residual calculation based on the operating data set; obtaining a critical abnormality situation of the long-term trend component, a critical abnormality situation of the periodic component and a critical abnormality situation of the residual term based on a preset growth analysis algorithm, a long-term trend component, a critical abnormality situation of the periodic component and a critical abnormality situation of the residual term; based on the critical abnormality situation of the long-term trend component, the critical abnormality situation of the periodic component and the critical abnormality situation of the residual term, performing abnormal monitoring on the operating data set through a random forest algorithm and a preset time window to obtain a potential fault monitoring result.

[0020] In the above solution, the corresponding algorithm is used to decompose the long-term trend, periodicity and residual components of the network topology operation data set, and then the growth analysis algorithm is used to quantify the critical anomaly threshold of each component. Combined with the random forest algorithm, multi-dimensional anomaly features are integrated within the time window to accurately identify potential failure modes and trigger conditions. This enables the disaster recovery strategy to dynamically adapt and adjust to differentiated failure scenarios such as trend offsets, periodic fluctuations and residual disturbances, thereby improving the accuracy and scenario coverage of anomaly detection, enhancing the adaptability of the target disaster recovery solution to different failure scenarios, and improving the reliability of business failure recovery.

[0021] Furthermore, when a business failure occurs, the target disaster recovery strategy is automatically switched based on the preset cloud policy library to recover the business failure, including: obtaining the network topology fault indicator based on the enterprise IT architecture operation indicators and potential fault monitoring results; obtaining the comprehensive health score based on the network topology fault indicator; if the comprehensive health score meets the preset switching threshold, the policy switch is triggered; if the policy switch is triggered, the resource fitness score is calculated based on the preset cloud policy library, and the target disaster recovery strategy whose resource fitness score meets the preset score requirements is automatically switched to recover the business failure.

[0022] In the above scheme, the network topology fault indicators are analyzed in real time to generate a comprehensive health score. When the score triggers the preset switching threshold, the fitness score of each disaster recovery strategy in the cloud policy library to the current resource status is calculated. Thus, the dual quantitative decision of health score and resource fitness score is adopted to realize automatic switching of target disaster recovery strategy, and then the fault characteristics are dynamically associated with resource status to ensure that the switched target disaster recovery strategy accurately matches the system recovery requirements under the real-time fault scenario. It avoids the problems of lag and subjectivity of traditional disaster recovery drills relying on manual intervention, breaks through the limitation of static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business fault scenarios, and improves the reliability of business fault recovery.

[0023] An embodiment of the present invention also provides a business fault recovery system based on disaster recovery strategy optimization, including: an enterprise IT architecture analysis module, an initial disaster recovery strategy generation module, a target disaster recovery strategy generation module and a fault recovery module; the enterprise IT architecture analysis module is used to obtain a network topology map and an enterprise digital twin model based on the enterprise IT architecture; the initial disaster recovery strategy generation module is used to obtain an initial disaster recovery strategy based on the network topology map; the target disaster recovery strategy generation module is used to input the initial disaster recovery strategy into the enterprise digital twin model, optimize the initial disaster recovery strategy until the preset convergence requirements are met, obtain the target disaster recovery strategy and store the target disaster recovery strategy in a preset cloud policy library; the fault recovery module is used to automatically switch the target disaster recovery strategy based on the preset cloud policy library when a business fault occurs to recover the business fault.

[0024] An embodiment of the present invention proposes a business fault recovery system based on disaster recovery strategy optimization. The linkage mechanism between the enterprise digital twin model and the network topology map is constructed through the enterprise IT architecture analysis module and the initial disaster recovery strategy generation module. The initial disaster recovery strategy is input into the enterprise digital twin model through the target disaster recovery strategy generation module for simulation iterative optimization. The real-time mapping and prediction capabilities of the enterprise digital twin model are utilized to make the target disaster recovery strategy deeply adapt to the risk characteristics of different business scenarios. At the same time, the target disaster recovery strategy is stored in a preset cloud policy library through the fault recovery module. When a fault occurs, intelligent matching and automatic switching of multi-dimensional disaster recovery plans are realized to recover business faults, breaking through the limitations of traditional static plans. Therefore, the reliability of business fault recovery is improved through continuous optimization of virtual-reality interaction and adaptive adjustment of cloud strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic flow chart of the steps of a service failure recovery method based on disaster recovery strategy optimization provided by an embodiment of the present invention;

[0026] Figure 2 A schematic diagram of the module structure of a business failure recovery system based on disaster recovery strategy optimization provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] Example 1

[0029] See also Figure 1 , Figure 1A flowchart of a business failure recovery method based on disaster recovery strategy optimization is provided in one embodiment of the present invention. Figure 1 As shown, the embodiment of the present invention proposes a service failure recovery method based on disaster recovery strategy optimization, including steps 101 to 104, each of which is specifically as follows:

[0030] Step 101: Based on the enterprise IT architecture, obtain a network topology map and an enterprise digital twin model;

[0031] As a preferred solution, executing step 101 includes: setting up device nodes based on the enterprise IT architecture and connecting the device nodes to obtain a network topology diagram; constructing a resource load change model based on the first resource load parameters of each device node; constructing a network communication model based on the second resource load parameters of each device node; and generating an enterprise digital twin model corresponding to the network topology diagram based on the resource load change model and the network communication model.

[0032] A preferred implementation method collects information about each network device in the enterprise IT architecture server and storage system, builds network links between each network device, and thereby establishes a network topology diagram, wherein a number of device nodes are set in the network topology diagram, each device node represents a different network device, and the connection between the nodes represents the network link between the network devices, thereby adding the business application corresponding to each network device to the corresponding device node. In this embodiment, the network link between the network devices can be constructed using existing technology; after obtaining the network topology diagram, the CPU utilization rate, memory utilization rate, and disk storage occupancy rate of each device (equivalent to the first resource load parameter) are counted, and a corresponding resource load change model is established based on each set of statistical data, and then the current network bandwidth utilization rate and network delay (equivalent to the second resource load parameter) of each network device are calculated to establish a network communication model, wherein the calculation formulas for network bandwidth utilization rate and network delay are as follows:

[0033]

[0034]

[0035] Where U NET Represents network bandwidth utilization; B used Represents the currently used bandwidth; B total Represents the total bandwidth; D la represents data transmission delay; H represents data transmission volume; B availRepresents the current available bandwidth; in this embodiment, the model construction process can be implemented by conventional technical means in this field; then the resource load change model and the network communication model are coupled to generate an enterprise digital twin model corresponding to the network topology diagram. In this embodiment, there are many ways to couple models, such as event-driven coupling mechanism, data interface coupling mechanism and model embedded coupling mechanism, etc., which can be adjusted according to actual application conditions.

[0036] In the above scheme, device nodes are first set up for the enterprise IT architecture and a network topology diagram is constructed. Then, the resource load change model established by the first resource load parameter and the network communication model generated by the second resource load parameter are combined to form an enterprise digital twin model with real-time mapping and prediction capabilities. The initial disaster recovery strategy is dynamically simulated and iteratively optimized through the enterprise digital twin model, so that the strategy can accurately match the resource load status and network communication characteristics of different device nodes, thereby adaptively adjusting the disaster recovery plan under business failure scenarios to recover business failures and improve the reliability of business failure recovery.

[0037] In order to improve the reliability of the enterprise digital twin model, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, which also includes: obtaining real-time device node data based on the network topology diagram; inputting the real-time device node data into the enterprise digital twin model to perform disaster scenario simulation drills to obtain the fault propagation situation; executing the preset disaster recovery strategy based on the fault propagation situation and calculating the simulation error; based on the simulation error, adjusting the model parameters of the enterprise digital twin model to optimize the enterprise digital twin model.

[0038] In one preferred implementation, the enterprise digital twin model receives real-time device node data generated by the network topology map and synchronously updates the model's status information. Disaster scenario simulation drills are performed based on preset disaster scenarios, such as server downtime, network link rupture, and storage device damage. Based on the propagation of the fault, a preset disaster recovery strategy (usually an automatically optimized disaster recovery strategy developed by the user) is executed to evaluate the disaster recovery time and resource allocation ratio, generate simulation prediction data, and then calculate the simulation error based on the actual system measurement data and the simulation prediction data. Finally, the parameters of the enterprise digital twin model are recalculated and adjusted based on the simulation error to optimize the enterprise digital twin model. The calculation formula for the simulation error is as follows:

[0039]

[0040] Where, E s represents the simulation error; M r Represents real system measurement data; M s Represents the simulation system prediction data.

[0041] In the above scheme, the device node data in the network topology is collected in real time and input into the enterprise digital twin model, dynamically simulating the fault propagation path and impact range in the disaster scenario. Then, based on the error analysis between the simulation results of the preset disaster recovery strategy and the actual fault propagation situation, the resource load and communication parameters of the digital twin model are reversely iterated and adjusted to make the model continuously approach the real business environment. Thus, through the closed-loop feedback mechanism of disaster recovery drills and parameter optimization, the prediction accuracy of the digital twin model for different disaster scenarios is enhanced, thereby improving the adaptability of the disaster recovery strategy to dynamic fault characteristics, realizing policy adaptive optimization driven by real-time data, and significantly improving the reliability of business fault recovery.

[0042] Step 102: Obtain an initial disaster recovery strategy based on the network topology diagram;

[0043] As a preferred solution, an initial disaster recovery strategy is obtained based on a network topology diagram, including: real-time monitoring of the network topology diagram, obtaining real-time disaster recovery data and acquiring characteristic information corresponding to the real-time disaster recovery data; constructing a state space based on the first characteristic information; constructing an action space based on the second characteristic information; based on the state space, calculating the initial probability of selecting a random action in the action space under any state; and obtaining an initial disaster recovery strategy based on the initial probability.

[0044] A preferred implementation method is to monitor the network topology in real time, obtain real-time disaster recovery data including: system log data, sensor monitoring data, historical disaster recovery plans and effects, and external environment information, etc., pre-process the real-time disaster recovery data and then perform standardization processing to obtain feature information corresponding to the real-time disaster recovery data, wherein the feature information includes: server status, network status, trigger switching, resource scheduling, and data loss amount, etc. In this embodiment, pre-processing and standardization can be implemented using technical means commonly used in this field; then, a state space is constructed based on the server status, network status, service availability, and fault type (equivalent to the first feature information), and an action space is established based on trigger switching, resource scheduling, and data recovery (equivalent to the second feature information). Finally, the initial probability of selecting a random action in the action space under any state in the state space is calculated, and the initial disaster recovery strategy is designed based on the distribution of the initial probability. In this embodiment, the calculation formula of the initial probability is as follows:

[0045]

[0046] Where π0(A|S) represents the initial probability of selecting a random action A in the action space at any state S in the state space; |A| represents the total number of optional actions in the action space;

[0047] In the above scheme, the disaster recovery data in the network topology diagram is monitored in real time, the first characteristic information is extracted to construct a state space to characterize the real-time operating status of the system, and the action space is generated according to the second characteristic information to define a set of executable disaster recovery operations. By calculating the initial probability distribution of random actions in the action space under each state in the state space, an initial disaster recovery strategy that dynamically adapts to the current system state is formed, so that the strategy generation process is closely related to the real-time environmental characteristics to achieve adaptive adjustment of the disaster recovery plan and improve the reliability of business fault recovery.

[0048] Step 103: Input the initial disaster recovery strategy into the enterprise digital twin model, optimize the initial disaster recovery strategy until it meets the preset convergence requirements, obtain the target disaster recovery strategy, and store the target disaster recovery strategy in a preset cloud strategy library;

[0049] As a preferred solution, the initial disaster recovery strategy is input into the enterprise digital twin model, the initial disaster recovery strategy is optimized until the preset convergence requirements are met, the target disaster recovery strategy is obtained and the target disaster recovery strategy is stored in the preset cloud strategy library, including: setting a reward function based on the third characteristic information; inputting the initial disaster recovery strategy into the enterprise digital twin model to perform a simulation drill, calculating the instant reward for selecting the corresponding action under each drill state through the reward function, and obtaining a first instant reward; adjusting the initial disaster recovery strategy based on the preset random disturbance, recalculating the instant reward for selecting the corresponding action under each drill state, and obtaining a second instant reward; based on the first instant reward and the second instant reward, optimizing the initial disaster recovery strategy until the preset convergence requirements are met, obtaining the target disaster recovery strategy, and storing the target disaster recovery strategy in the preset cloud strategy library.

[0050] In a preferred implementation method, a reward function is first set based on minimizing the service interruption time and data loss amount (equivalent to the third characteristic information). The expression of the reward function is as follows:

[0051] R=-(λ1T re +λ2L da );

[0052] In the formula, R represents the immediate reward; λ1 and λ2 represent weighting parameters; T re L stands for recovery time; daRepresents the amount of data loss; the initial disaster recovery strategy is then input into the enterprise digital twin model to perform a simulation drill, and the instant reward for selecting the corresponding action under each drill state is calculated through the reward function. This instant reward is used as the first instant reward, and the current initial disaster recovery strategy is adjusted by adding random perturbations. The instant reward for selecting the corresponding action under each drill state is recalculated through the reward function. This instant reward is used as the second instant reward, and then the Q value (action value function) between the first instant reward and the second instant reward is calculated to optimize the initial disaster recovery strategy until the preset convergence requirements are met, and the target disaster recovery strategy is obtained, and the target disaster recovery strategy is stored in the preset cloud strategy library.

[0053] In the above scheme, a reward function is first constructed using the third characteristic information. When simulating the initial disaster recovery strategy in the enterprise digital twin model, the reward function is used to quantify the first instant rewards of different actions in the drill state, and the second instant reward is generated through a random perturbation strategy. Then, the initial disaster recovery strategy is adjusted according to the first instant reward and the second instant reward. The reinforcement learning mechanism is used to drive the iterative optimization of the disaster recovery strategy in the virtual environment until convergence, so that the initial disaster recovery strategy is deeply adapted to the complex fault scenario characteristics simulated by the digital twin model. Finally, the target disaster recovery strategy is generated and stored in the preset cloud policy library, forming an adaptive strategy generation closed loop based on real-time drill feedback, breaking through the limitation of the static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business failure scenarios, and improving the reliability of business fault recovery.

[0054] In the above preferred scheme, based on the first instant reward and the second instant reward, the initial disaster recovery strategy is optimized until the preset convergence requirements are met, the target disaster recovery strategy is obtained, and the target disaster recovery strategy is stored in the preset cloud strategy library, including: based on the preset reinforcement learning algorithm and the first instant reward, the first action value function is obtained; based on the preset reinforcement learning algorithm and the second instant reward, the second action value function is obtained; based on the second action value function, the initial disaster recovery strategy is updated, and the change value between the first action value function and the second action value function is calculated; if the change value between the first action value function and the second action value function meets the preset convergence requirements, the updated initial disaster recovery strategy is used as the target disaster recovery strategy, and the target disaster recovery strategy is stored in the preset cloud strategy library.

[0055] A preferred implementation method is to calculate the Q value (action value function) between the first instant reward and the second instant reward before and after the initial disaster recovery strategy is adjusted by the DQN algorithm. In this embodiment, the action value function calculated by the first instant reward is equivalent to the first action value function, and the action value function calculated by the second instant reward is equivalent to the second action value function. At the same time, according to the Q value (equivalent to the second action value function) calculated after the initial disaster recovery strategy is adjusted, the action with the highest Q value (equivalent to the second action value function) in each state is selected, the initial disaster recovery strategy is updated, and the Q value change before and after the initial disaster recovery strategy is updated is calculated. If the change value is less than the preset threshold, it is considered that the strategy converges and the initial disaster recovery strategy update adjustment is stopped, and the target disaster recovery strategy is output, wherein the preset threshold can be adjusted according to the actual application situation; otherwise, the initial disaster recovery strategy continues to be updated and adjusted, and finally the output target disaster recovery strategy is stored in the preset cloud strategy library. In this embodiment, the Q value calculation method is as follows:

[0056]

[0057] Where, Q(S t ,A t ) represents the Q value of state S selecting action A at time t; R t represents the immediate reward at time t; γ represents the discount factor; Represents the maximum Q value of all actions A′ in state S at time t+1.

[0058] In the above scheme, a preset reinforcement learning algorithm is used to construct the first action value function and the second action value function based on the first immediate reward and the second immediate reward respectively. The optimization direction of the disaster recovery strategy is dynamically evaluated by comparing the change values ​​of the two, and the initial disaster recovery strategy is iteratively updated in the digital twin environment. When the change values ​​of the first action value function and the second action value function meet the convergence conditions, it indicates that the decision-making benefits of the strategy under the simulated fault scenario tend to be stable. At this time, the updated initial disaster recovery strategy is used as the target disaster recovery strategy and stored in the preset cloud policy library, forming an adaptive strategy generation closed loop based on real-time exercise feedback, breaking through the limitation of the static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business failure scenarios, and improving the reliability of business fault recovery.

[0059] In order to ensure the reliability of the target disaster recovery strategy, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, which also includes: obtaining actual drill results based on the target disaster recovery strategy; inputting the target disaster recovery strategy into the enterprise digital twin model to perform simulation drills to obtain simulation results; obtaining a drill error based on the actual drill results and the simulation results; if the drill error does not meet the preset threshold requirements, optimizing the target disaster recovery strategy.

[0060] As a preferred implementation method, the target disaster recovery strategy is applied to actual drills to obtain actual drill results, and then the target disaster recovery strategy is input into the enterprise digital twin model for simulation drills to obtain simulation results. During the simulation drills, simulation data can also be collected to calculate the optimized disaster recovery time and evaluate the adaptability of the target disaster recovery strategy.

[0061] After obtaining the actual drill results and simulation results, the drill error between the actual drill results and the simulation results is calculated. If the drill error exceeds the set threshold (equivalent to the preset threshold requirement), the target disaster recovery strategy is re-optimized and a new drill plan is executed to optimize the target disaster recovery strategy until the drill error meets the preset threshold requirement.

[0062] In the above scheme, the target disaster recovery strategy is input into the digital twin model to perform simulation exercises to generate simulation results, and the error is compared with the actual exercise results. When the error exceeds the preset threshold, the strategy iterative optimization is triggered. Then, by judging whether the exercise error between the actual exercise results and the simulation results meets the threshold requirements, the deviation between the digital twin model prediction and the actual fault recovery process is continuously eliminated, so that the target disaster recovery strategy dynamically approaches the optimal solution of the real business environment in the feedback loop of virtual and real integration, thereby enhancing the adaptability of the target disaster recovery strategy to different fault scenarios and improving the reliability of business fault recovery.

[0063] Before executing step 104, as a preferred solution of an embodiment of the present invention, an embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, including: obtaining an operating data set based on a network topology diagram; obtaining a long-term trend component based on a preset local regression algorithm and the operating data set; obtaining a periodic component based on a preset fitting algorithm and the operating data set; obtaining a residual term through residual calculation based on the operating data set; obtaining a critical abnormality situation of the long-term trend component, a critical abnormality situation of the periodic component, and a critical abnormality situation of the residual term based on a preset growth analysis algorithm, a long-term trend component, a critical abnormality situation of the periodic component, and a critical abnormality situation of the residual term; based on the critical abnormality situation of the long-term trend component, the critical abnormality situation of the periodic component, and the critical abnormality situation of the residual term, performing abnormal monitoring on the operating data set through a random forest algorithm and a preset time window to obtain a potential fault monitoring result.

[0064] As a preferred implementation method, network device operation data is collected in the network topology diagram. The network device operation data includes: CPU usage, memory usage, disk I / O rate, network traffic and service response time, etc., and then each group of data is preprocessed and normalized. The preprocessing includes: data cleaning and denoising. The preprocessing and normalization can be achieved by using existing technologies. Each group of operation data is integrated into an operation data set. The processed operation data set is locally regressed through LOESS smoothing (equivalent to a preset local regression algorithm) to obtain the long-term trend component in the operation data set. The moving average (equivalent to a preset fitting algorithm) is used to filter, separate and extract the periodic component in the operation data set. The residual term in the operation data set is extracted through residual calculation.

[0065] After decomposing the operating data of each network device in the network topology diagram, STL component analysis (equivalent to the preset growth analysis algorithm) is performed on the long-term trend component, cyclical component, and residual term. If the long-term trend component continues to decline (equivalent to the critical abnormality of the long-term trend component), it indicates that the system health is deteriorating. If the change value of the cyclical component reaches the preset abnormal threshold (equivalent to the critical abnormality of the cyclical component), it indicates that the cyclical load pattern is abnormal. If the residual term is far away from zero (equivalent to the critical abnormality of the residual term), it indicates a potential fault.

[0066] After analyzing the critical conditions of long-term trend component anomalies, periodic component anomalies, and residual anomalies, the random forest algorithm is used to randomly select multiple groups of sample subsets from the running data set, and use each selected sample subset as the root node to build a corresponding number of binary trees. For each binary tree, a feature point in the sample subset is randomly selected, and a split point is randomly selected on the selected feature point as the partition threshold. Then, each sample subset is divided into two left and right subsets according to the partition threshold. Feature point selection, partition threshold setting, and subset partitioning are repeated for each newly generated subset. When each group of subsets after division contains only one group of samples, the partitioning is stopped. The number of splits of each sample subset is counted, and the statistical result is used as the path length of each binary tree. Then calculate the average path length of all binary trees and normalize the calculation results. If the normalized result is higher than the preset length threshold, the corresponding operating data is determined to be an anomaly point, record all abnormal samples, set a group of time windows (equivalent to the preset time windows), and calculate the proportion of anomalies in time windows of different time periods based on the critical abnormal conditions of long-term trend components, the critical abnormal conditions of periodic components, and the critical abnormal conditions of residual items. The future trend is predicted based on the proportion of anomalies in each time window, and then the time periods where the abnormal proportion exceeds the preset fault threshold are extracted, that is, the time of occurrence of future faults as the potential fault monitoring result. In this embodiment, the setting of each preset value needs to be set according to the actual application situation, and can also be set based on historical analysis experience.

[0067] In the above solution, the corresponding algorithm is used to decompose the long-term trend, periodicity and residual components of the network topology operation data set, and then the growth analysis algorithm is used to quantify the critical anomaly threshold of each component. Combined with the random forest algorithm, multi-dimensional anomaly features are integrated within the time window to accurately identify potential failure modes and trigger conditions. This enables the disaster recovery strategy to dynamically adapt and adjust to differentiated failure scenarios such as trend offsets, periodic fluctuations and residual disturbances, thereby improving the accuracy and scenario coverage of anomaly detection, enhancing the adaptability of the target disaster recovery solution to different failure scenarios, and improving the reliability of business failure recovery.

[0068] Step 104: When a business failure occurs, the target disaster recovery strategy is automatically switched based on the preset cloud strategy library to recover the business failure.

[0069] As an optimal solution, when a business failure occurs, the target disaster recovery strategy is automatically switched based on the preset cloud policy library to recover the business failure, including: obtaining the network topology fault indicator based on the enterprise IT architecture operation indicators and potential fault monitoring results; obtaining the comprehensive health score based on the network topology fault indicator; if the comprehensive health score meets the preset switching threshold, the policy switch is triggered; if the policy switch is triggered, the resource fitness score is calculated based on the preset cloud policy library, and the target disaster recovery strategy whose resource fitness score meets the preset score requirements is automatically switched to recover the business failure.

[0070] In a preferred implementation method, the enterprise IT architecture operation indicators and potential fault monitoring results in a multi-cloud environment are collected. The enterprise IT architecture operation indicators include key indicators of business load indicators, service quality indicators, and multi-cloud resource health status in the multi-cloud environment. The network topology fault indicators include the response time, packet loss rate, and error rate of each network device in the network topology. The comprehensive health score of the current target disaster recovery decision is calculated based on the network topology fault indicators. The specific calculation formula of the comprehensive health score is as follows:

[0071] S h (t) = w1Q r (t)+w2Q l (t)+w3Q E (t);

[0072] Where S h (t) represents the comprehensive health score; Q r (t) represents the response time; Q l (t) represents the packet loss rate; Q E (t) represents the error rate; w1, w2 and w3 represent weighting factors respectively;

[0073] If the comprehensive health score is higher than the preset switching threshold, the switch is triggered and the resource fitness score of each target disaster recovery decision in the preset cloud policy library is calculated based on the current load, health status and switching cost of the current target disaster recovery decision. When a disaster occurs, the target disaster recovery strategy with the highest resource fitness score (equivalent to the preset score requirement) is automatically switched to recover the business failure. The calculation formula of the resource fitness score is as follows:

[0074] R i (t) = αL i (t)+βH i (t)+γC i (t);

[0075] Where R i (t) represents the resource fitness score of the i-th group of cloud services; L i (t) represents the current load of the i-th group of cloud services; H i (t) represents the health status of the i-th group of cloud services; C i (t) represents the switching cost of the i-th group of cloud services; α, β, and γ represent weight coefficients respectively;

[0076] In order to ensure the ability to quickly respond to various disaster scenarios, the target disaster recovery strategy is automatically switched according to the resource fitness score. Then, the selected target disaster recovery strategy is pointed to by modification and parsing, and the traffic routing is updated. The database and cache consistency are checked, and the comprehensive health score of the target disaster recovery strategy is recalculated. If the comprehensive health score is lower than the preset switching threshold, the target disaster recovery strategy will not be switched. Otherwise, the target disaster recovery strategy that meets the preset score requirements will be automatically switched based on the resource fitness score.

[0077] In the above scheme, the network topology fault indicators are analyzed in real time to generate a comprehensive health score. When the score triggers the preset switching threshold, the fitness score of each disaster recovery strategy in the cloud policy library to the current resource status is calculated. Thus, the dual quantitative decision of health score and resource fitness score is adopted to realize automatic switching of target disaster recovery strategy, and then the fault characteristics are dynamically associated with resource status to ensure that the switched target disaster recovery strategy accurately matches the system recovery requirements under the real-time fault scenario. It avoids the problems of lag and subjectivity of traditional disaster recovery drills relying on manual intervention, breaks through the limitation of static plan configuration mechanism that is difficult to adaptively adjust the disaster recovery plan according to different business fault scenarios, and improves the reliability of business fault recovery.

[0078] As another example of an embodiment of the present invention, a fault warning function can also be set in the enterprise IT architecture. Whenever an abnormal trend appears in the enterprise IT architecture, a warning signal is issued to enable the enterprise IT architecture to quickly respond to the fault and automatically switch the target disaster recovery strategy to repair the fault. In addition, for each enterprise digital twin model drill, the success rate, execution time and resource consumption of the disaster recovery drill can be recorded, and the consistency of the drill results with the plan can be verified to generate an audit report. According to the audit report, various preset thresholds can be adjusted to enable the enterprise IT architecture to achieve intelligent matching and automatic switching of multi-dimensional disaster recovery solutions.

[0079] An embodiment of the present invention proposes a business fault recovery method based on disaster recovery strategy optimization, constructs a linkage mechanism between the enterprise digital twin model and the network topology map, inputs the initial disaster recovery strategy into the enterprise digital twin model for simulation iterative optimization, and utilizes the real-time mapping and prediction capabilities of the enterprise digital twin model to make the target disaster recovery strategy deeply adapt to the risk characteristics of different business scenarios. At the same time, the target disaster recovery strategy is also stored in a preset cloud policy library. When a fault occurs, intelligent matching and automatic switching of multi-dimensional disaster recovery plans are realized to recover business faults, breaking through the limitations of traditional static plans. Therefore, the reliability of business fault recovery is improved through continuous optimization of virtual-reality interaction and adaptive adjustment of cloud strategies.

[0080] Example 2

[0081] See also Figure 2 , Figure 2 The module structure diagram of a business failure recovery system based on disaster recovery strategy optimization provided by an embodiment of the present invention. Figure 2 As shown, an embodiment of the present invention also provides a business fault recovery system based on disaster recovery strategy optimization, including: an enterprise IT architecture analysis module 201, an initial disaster recovery strategy generation module 202, a target disaster recovery strategy generation module 203 and a fault recovery module 204; the enterprise IT architecture analysis module 201 is used to obtain a network topology map and an enterprise digital twin model based on the enterprise IT architecture; the initial disaster recovery strategy generation module 202 is used to obtain an initial disaster recovery strategy based on the network topology map; the target disaster recovery strategy generation module 203 is used to input the initial disaster recovery strategy into the enterprise digital twin model, optimize the initial disaster recovery strategy until the preset convergence requirements are met, obtain the target disaster recovery strategy and store the target disaster recovery strategy in a preset cloud policy library; the fault recovery module 204 is used to automatically switch the target disaster recovery strategy based on the preset cloud policy library when a business fault occurs to recover the business fault.

[0082] An embodiment of the present invention proposes a business fault recovery system based on disaster recovery strategy optimization. The linkage mechanism between the enterprise digital twin model and the network topology map is constructed through the enterprise IT architecture analysis module and the initial disaster recovery strategy generation module. The initial disaster recovery strategy is input into the enterprise digital twin model through the target disaster recovery strategy generation module for simulation iterative optimization. The real-time mapping and prediction capabilities of the enterprise digital twin model are utilized to make the target disaster recovery strategy deeply adapt to the risk characteristics of different business scenarios. At the same time, the target disaster recovery strategy is stored in a preset cloud policy library through the fault recovery module. When a fault occurs, intelligent matching and automatic switching of multi-dimensional disaster recovery plans are realized to recover business faults, breaking through the limitations of traditional static plans. Therefore, the reliability of business fault recovery is improved through continuous optimization of virtual-reality interaction and adaptive adjustment of cloud strategies.

[0083] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0084] In the description of this specification, the reference terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0085] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, features specified as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of this application, "plurality" means two or more, unless otherwise specifically defined.

Claims

1. A business failure recovery method based on disaster recovery strategy optimization, characterized in that: include: Based on the enterprise IT architecture, obtain the network topology diagram and enterprise digital twin model; Based on the network topology diagram, an initial disaster recovery strategy is obtained; Inputting the initial disaster recovery strategy into the enterprise digital twin model, optimizing the initial disaster recovery strategy until a preset convergence requirement is met, obtaining a target disaster recovery strategy, and storing the target disaster recovery strategy in a preset cloud strategy library; When a business failure occurs, the target disaster recovery strategy is automatically switched based on the preset cloud strategy library to recover the business failure.

2. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 1, characterized in that: Based on the enterprise IT architecture, the network topology diagram and enterprise digital twin model are obtained, including: Setting device nodes based on the enterprise IT architecture and connecting the device nodes to obtain a network topology diagram; Building a resource load change model based on the first resource load parameter of each of the device nodes; Building a network communication model based on the second resource load parameter of each of the device nodes; An enterprise digital twin model corresponding to the network topology diagram is generated based on the resource load change model and the network communication model.

3. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 2, characterized in that: Also includes: Based on the network topology diagram, real-time device node data is obtained; Inputting the real-time device node data into the enterprise digital twin model to conduct a disaster scenario simulation drill to obtain fault propagation conditions; Execute a preset disaster recovery strategy based on the fault propagation situation and calculate a simulation error; Based on the simulation error, model parameters of the enterprise digital twin model are adjusted to optimize the enterprise digital twin model.

4. The method for business failure recovery based on disaster recovery strategy optimization according to claim 1, characterized in that: Based on the network topology diagram, an initial disaster recovery strategy is obtained, including: Monitor the network topology in real time, obtain real-time disaster recovery data and acquire feature information corresponding to the real-time disaster recovery data; Constructing a state space based on the first feature information; constructing an action space based on the second feature information; Based on the state space, calculating the initial probability of selecting a random action in the action space in any state; Based on the initial probability, an initial disaster recovery strategy is obtained.

5. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 4, characterized in that: Inputting the initial disaster recovery strategy into the enterprise digital twin model, optimizing the initial disaster recovery strategy until a preset convergence requirement is met, obtaining a target disaster recovery strategy, and storing the target disaster recovery strategy in a preset cloud strategy library, including: Setting a reward function based on the third feature information; Inputting the initial disaster recovery strategy into the enterprise digital twin model to perform a simulation exercise, and calculating the instantaneous reward for selecting a corresponding action under each exercise state using the reward function to obtain a first instantaneous reward; Adjusting the initial disaster recovery strategy based on a preset random disturbance, recalculating the instantaneous reward for selecting the corresponding action in each drill state, and obtaining a second instantaneous reward; Based on the first instant reward and the second instant reward, the initial disaster recovery strategy is optimized until a preset convergence requirement is met, a target disaster recovery strategy is obtained, and the target disaster recovery strategy is stored in a preset cloud strategy library.

6. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 5, characterized in that: Based on the first instant reward and the second instant reward, optimizing the initial disaster recovery strategy until a preset convergence requirement is met, obtaining a target disaster recovery strategy, and storing the target disaster recovery strategy in a preset cloud strategy library, including: Obtaining a first action-value function based on a preset reinforcement learning algorithm and the first immediate reward; Obtaining a second action-value function based on the preset reinforcement learning algorithm and the second immediate reward; Updating the initial disaster recovery strategy based on the second action-value function, and calculating a change value between the first action-value function and the second action-value function; If the change value between the first action value function and the second action value function meets the preset convergence requirement, the updated initial disaster recovery strategy is used as the target disaster recovery strategy, and the target disaster recovery strategy is stored in the preset cloud strategy library.

7. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 5 or 6, characterized in that: Also includes: Based on the target disaster recovery strategy, obtain actual drill results; Inputting the target disaster recovery strategy into the enterprise digital twin model to perform simulation exercises and obtain simulation results; Obtaining a drill error based on the actual drill result and the simulation result; If the drill error does not meet the preset threshold requirement, the target disaster recovery strategy is optimized.

8. A method for recovering a business failure based on disaster recovery strategy optimization according to any one of claims 1 to 6, characterized in that: include: Based on the network topology diagram, an operation data set is obtained; obtaining a long-term trend component based on a preset local regression algorithm and the running data set; acquiring a periodic component based on a preset fitting algorithm and the running data set; Based on the running data set, a residual term is obtained by residual calculation; Based on a preset growth analysis algorithm, the long-term trend component, the periodic component and the residual term, obtaining a critical abnormality of the long-term trend component, a critical abnormality of the periodic component and a critical abnormality of the residual term; Based on the critical abnormal conditions of the long-term trend component, the critical abnormal conditions of the periodic component and the critical abnormal conditions of the residual term, the operation data set is monitored for abnormalities using a random forest algorithm and a preset time window to obtain potential fault monitoring results.

9. A method for recovering a business failure based on disaster recovery strategy optimization according to claim 8, characterized in that: When a business failure occurs, automatically switching the target disaster recovery strategy based on the preset cloud strategy library to recover the business failure includes: Based on the enterprise IT architecture operation indicators and the potential fault monitoring results, a network topology fault indicator is obtained; Based on the fault indicators of the network topology, a comprehensive health score is obtained; If the comprehensive health score meets the preset switching threshold, a policy switch is triggered; If policy switching is triggered, the resource fitness score is calculated based on the preset cloud policy library, and the target disaster recovery policy whose resource fitness score meets the preset score requirements is automatically switched to recover the business failure.

10. A business failure recovery system based on disaster recovery strategy optimization, characterized in that: Executing a service failure recovery method based on disaster recovery strategy optimization according to any one of claims 1 to 9, comprising: Enterprise IT architecture analysis module, initial disaster recovery strategy generation module, target disaster recovery strategy generation module and fault recovery module; The enterprise IT architecture analysis module is used to obtain a network topology diagram and an enterprise digital twin model based on the enterprise IT architecture; The initial disaster recovery strategy generation module is used to obtain an initial disaster recovery strategy based on the network topology diagram; The target disaster recovery strategy generation module is used to input the initial disaster recovery strategy into the enterprise digital twin model, optimize the initial disaster recovery strategy until the preset convergence requirements are met, obtain the target disaster recovery strategy, and store the target disaster recovery strategy in a preset cloud strategy library; The fault recovery module is used to automatically switch the target disaster recovery strategy based on the preset cloud strategy library when a business failure occurs, so as to recover the business failure.

Images