Server cluster abnormal scene repair task generation method based on big data AI

Abstract

The invention discloses a server cluster abnormal scene repair task generation method based on big data AI, and relates to the technical field of machine learning, and the method comprises the steps: obtaining a historical operation log and real-time operation data of a server cluster, analyzing the multi-mode abnormal fusion of the interaction relation between a hardware layer and a software layer of the server cluster, and generating a repair task of the abnormal scene of the server cluster; generating abnormal operation scene behavior data of the target server cluster component; performing causal association analysis according to the abnormal operation scene behavior data of the target server cluster component and the known server cluster abnormal scene behavior performance big data, and evaluating the current abnormal deviation state of the target server cluster component; and based on the current abnormal deviation state of the target server cluster component, marking an abnormal component node and a corresponding associated task sequence, establishing an uninterrupted task migration resource allocation model, and generating a server cluster abnormal scene repair task. The method has the advantages that the anomaly detection accuracy is improved, the fault recovery time is shortened, and the recovery capability of the server cluster is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and specifically relates to a method for generating abnormal scenario repair tasks for a server cluster based on big data AI. Background Art

[0002] Existing abnormal scenario repair solutions for server clusters based on big data AI implement anomaly detection through algorithms such as Isolation Forest and K-Means, and combine simple threshold alarms and predefined scripts to execute repair tasks. This leads to problems such as hardware-software data fragmentation, lack of causal reasoning, and rigidity of repair strategies during the actual repair process of server cluster abnormal scenarios, resulting in a high false alarm rate, a relatively long average repair time, and difficulty in handling cross-component cascading failures and dynamic resource competition scenarios. Secondary anomalies or service interruptions are likely to occur during the repair process. Summary of the Invention

[0003] To solve the above technical problems, a method for generating abnormal scenario repair tasks for a server cluster based on big data AI is provided. This technical solution solves the problems of the existing abnormal scenario repair solutions for server clusters based on big data AI that implement anomaly detection through algorithms such as Isolation Forest and K-Means, and combine simple threshold alarms and predefined scripts to execute repair tasks. This leads to problems such as hardware-software data fragmentation, lack of causal reasoning, and rigidity of repair strategies during the actual repair process of server cluster abnormal scenarios, resulting in a high false alarm rate, a relatively long average repair time, and difficulty in handling cross-component cascading failures and dynamic resource competition scenarios. Secondary anomalies or service interruptions are likely to occur during the repair process.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for generating abnormal scenario repair tasks for a server cluster based on big data AI, including: S1. Obtain the historical operation logs and real-time operation data of the server cluster, analyze the multi-modal anomaly fusion of the interaction relationship between the hardware layer and the software layer of the server cluster, and generate abnormal scenario behavior data for the operation of the target server cluster components; S2. Perform causal association analysis on the abnormal scenario behavior data of the target server cluster component operation and the big data of the known abnormal scenario behavior performance of the server cluster, and evaluate the current abnormal deviation state of the target server cluster component; S3. Based on the current abnormal deviation state of the target server cluster component, mark the abnormal component nodes and the corresponding associated task sequences, establish a non-interruptible task migration resource allocation model, and generate abnormal scenario repair tasks for the server cluster.

[0005] Preferably, according to the unit timestamp, the original performance parameters of each component of the server cluster are collected, and the original performance parameters of each component are preprocessed using sliding window normalization to form a hardware performance normalization vector of the server cluster; According to the unit timestamp, the historical operation logs of each component of the server cluster are collected, the BERT model is trained, the log semantic set of the server cluster is extracted and substituted into the Softmax function to assign weights to the log semantics of the server cluster; Using the attention weighting method, with the known abnormal operation scenarios of the server cluster components as the query matrix, the log semantic set as the query key, and the hardware performance normalization vector of the server cluster as the query value, the log feature vector of the server cluster is calculated.

[0006] Preferably, using the hardware performance normalization vector of the server cluster as the hardware node and the log feature vector of the server cluster as the software node, the Pearson correlation coefficient is used to calculate the correlation coefficient between the hardware node and the software node to generate a hardware node-software node connection edge, and the Sigmoid function is used to assign weights to the full connection edge of the hardware node-software node to establish a full connection edge network of the hardware node-software node of the server cluster; Extract each adjacent node in the full connection edge network of the hardware node-software node of the server cluster, establish an adjacent node array of the server cluster, and form a hardware node-software adjacent node matrix of the server cluster; Using the hardware node-software adjacent node matrix of the server cluster, train the GCN graph convolutional network, calculate the abnormal probability of each hardware node-software adjacent node using the fully connected layer, and form the behavior data of the abnormal operation scenarios of the target server cluster components.

[0007] Preferably, using the DTW dynamic event regularization algorithm, recursively calculate the maximum similarity between the big data of the abnormal operation scenario behavior performance of the known server cluster and the behavior data of the abnormal operation scenario of the target server cluster components, screen out the characteristic parameters of the abnormal operation scenario behavior performance of the known server cluster, and establish the characteristic time series data of the abnormal operation scenario performance of the known target server cluster components; Based on the characteristic time series data of the abnormal operation scenario performance of the known target server cluster components, substitute it into the full connection edge network of the hardware node-software node of the server cluster, and establish a causal graph of the abnormal operation scenario of the known target server cluster components according to the computing power call path of the hardware node-software node; Using the independent condition test method, verify the correlation strength of each edge of the causal graph of the abnormal operation scenario of the known target server cluster components, determine the causal chain between each node, and obtain the causal structure graph of the abnormal operation scenario of the known target server cluster components; Based on the MLP (Multi-Layer Perceptron), verify the influence condition expectation of the abnormal scenario operation performance characteristics corresponding to the numerical values of the hardware nodes and software nodes in the causal structure diagram of the abnormal scenario of the components of the known target server cluster, and obtain the hardware-software causal intervention characterization related to the abnormal scenario of the components of the known target server cluster.

[0008] Preferably, determine the hardware node-software node operation standardization parameters of the components of the known target server cluster, and form an array of standardization parameters of the components of the known target server cluster; Normalize the hardware-software causal intervention characterization related to the abnormal scenario of the components of the known target server cluster, and assign weights to the time series data of the abnormal scenario operation performance characteristics of the components of the known target server cluster; Using the Euclidean distance formula, with the weight of the abnormal scenario operation performance characteristics of the components of the known target server cluster as the external intervention factor, calculate the spatial distance between the time series data of the abnormal scenario operation performance characteristics of the components of the known target server cluster and the array of standardization parameters of the components of the known target server cluster group, and determine the current abnormal deviation status index of the components of the target server cluster;

[0009] Among them, is the current abnormal deviation status index of the i-th component of the target server cluster, is the weight of the j-th abnormal scenario operation performance characteristic of the i-th component of the known target server cluster, is the value of the j-th abnormal scenario operation performance characteristic of the i-th component of the known target server cluster, is the standardization parameter of the i-th component of the target server cluster, and n is the total number of components of the target server cluster.

[0010] Preferably, based on the currently to-be-executed associated task sequence of the components of the target server cluster, substitute it into the full-connected edge network of the hardware nodes-software nodes of the server cluster, determine the component load resource demand characteristic parameters of the to-be-executed associated task sequence, and form the current to-be-executed associated task demand characteristic vector of the components of the target server cluster; Using the full-connected edge network of the hardware nodes-software nodes of the server cluster, verify the resource conflict intensity between the current to-be-executed associated task demand characteristic vectors of the components of the target server cluster per unit time, and determine the priority of the current to-be-executed associated task demand of the components of the target server cluster.

[0011] Preferably, based on Logistic regression, with the current abnormal deviation status index of the components of the target server cluster as the given condition, with the component load demand characteristic data per unit time of the to-be-executed associated task sequence as the output, and with the shutdown risk probability of the components of the target server cluster executing the associated task as the label data output; Determine the initialization performance parameters of the target server cluster component according to the construction parameters of the target server cluster component, and establish a digital twin model of the target server cluster component. Update the digital twin model of the target server cluster component by using the shutdown risk probability of the associated tasks executed by the server cluster component, and determine the loadable performance support vector of the target server cluster component. Use the loadable performance support vector of the target server cluster component to train the SVR support vector machine. Taking the priority of the current associated task requirements to be executed by the target server cluster component as a limiting condition and maximizing the completion of the associated tasks to be executed as the objective function, generate the hyperplane boundary of the loadable performance execution task sequence to be executed by the target server cluster component, with the current demand feature vector of the associated tasks to be executed by the target server cluster component as the output and the abnormal scenario repair task of the server cluster as the output.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a solution for generating abnormal scenario repair tasks for server clusters based on big data AI. This solution realizes precise anomaly detection through multi-modal fusion analysis of the hardware-software interaction relationship, quantifies the abnormal deviation state by combining dynamic causal reasoning and intervention effect analysis, and uses digital twin simulation and SVR optimization to generate a lossless migration plan, forming a complete closed-loop of "anomaly perception - root cause analysis - intelligent repair", improving the anomaly detection accuracy, fault recovery time and repair ability of the server cluster compared with traditional methods. Brief Description of the Drawings

[0013] Figure 1 It is a flowchart of a method for generating abnormal scenario repair tasks for server clusters based on big data AI; Detailed Embodiments

[0014] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations.

[0015] Refer to Figure 1 As shown, the method for generating abnormal scenario repair tasks for server clusters based on big data AI includes: Step 1: Obtain the historical operation logs and real-time operation data of the server cluster, analyze the multi-modal anomaly fusion of the hardware layer and software layer interaction relationship of the server cluster, and generate the operation abnormal scenario behavior data of the target server cluster component. The content of the above step 1 includes the following: Collect the original performance parameters of each component of the server cluster according to the unit timestamp, preprocess the original performance parameters of each component using sliding window normalization, and form the hardware performance normalization vector of the server cluster; Collect the historical operation logs of each component of the server cluster according to the unit timestamp, train the BERT model, extract the log semantic set of the server cluster and substitute it into the Softmax function to assign weights to the log semantics of the server cluster; Using the attention weighting method, with the known abnormal operation scenarios of the server cluster components as the query matrix, the log semantic set as the query key, and the hardware performance normalization vector of the server cluster as the query value, calculate the log feature vector of the server cluster; Using the hardware performance normalization vector of the server cluster as the hardware node and the log feature vector of the server cluster as the software node, calculate the correlation coefficient between the hardware node and the software node using the Pearson correlation coefficient to generate the hardware node-software node connection edge, and use Sigmoid to assign weights to the full connection edge of the hardware node-software node to establish the hardware node-software node full connection edge network of the server cluster; Extract each adjacent node in the hardware node-software node full connection edge network of the server cluster, establish the adjacent node array of the server cluster, and form the hardware node-software adjacent node matrix of the server cluster; Using the hardware node-software adjacent node matrix of the server cluster, train the GCN graph convolutional network, calculate the abnormal probability of each hardware node-software adjacent node using the fully connected layer, and form the abnormal scenario behavior data of the target server cluster component operation; When in use, combine the content of the above steps, As a further content, the current server cluster monitoring mainly relies on threshold alarms and single-dimensional anomaly detection (such as Prometheus, Isolation Forest), while log analysis mostly uses keyword matching (such as the ELK stack), resulting in the fragmentation of hardware performance data and software log semantics and the lack of dynamic interaction relationships. Moreover, traditional graph neural networks only model static hardware topologies and cannot capture cross-level associated anomalies and root cause propagation paths, causing problems such as high false alarm rates and difficult root cause positioning.

[0016] This solution realizes the cross-modal dynamic association between hardware performance indicators and log semantics through Pearson coefficient and Sigmoid weighting, strengthens key anomaly features by combining BERT-Softmax and the attention mechanism, uses the GCN graph convolutional network to output an interpretable anomaly probability distribution and propagation path, and at the same time adopts sliding window normalization and lightweight adjacency matrix design to achieve accurate root cause positioning.

[0017] Step 2: Conduct a causal correlation analysis between the abnormal scenario behavior data of the target server cluster components and the big data of the known abnormal scenario behavior manifestations of the server cluster, and evaluate the current abnormal deviation state of the target server cluster components; The content of Step 2 is as follows: Using the DTW dynamic event alignment algorithm, recursively calculate the maximum similarity value between the big data of the known abnormal scenario behavior manifestations of the server cluster and the abnormal scenario behavior data of the target server cluster components, screen out the characteristic parameters of the known abnormal scenario behavior manifestations of the server cluster, and establish the characteristic time series data of the abnormal scenario operation performance of the known target server cluster components; As a further content, the abnormal scenario operation performance characteristics include but are not limited to: abnormal curves (such as CPU / memory metrics changing over time), root cause location (Redis connection pool exhaustion) Based on the characteristic time series data of the abnormal scenario operation performance of the known target server cluster components, substitute it into the full-connected edge network of the hardware nodes - software nodes of the server cluster, and establish a causal graph of the abnormal scenario operation of the known target server cluster components according to the computing power call path of the hardware nodes - software nodes; Using the independent condition test method, verify the association strength of each edge of the causal graph of the abnormal scenario operation of the known target server cluster components, determine the causal chain between each node, and obtain the causal structure diagram of the abnormal scenario operation of the known target server cluster components; Based on the MLP multi-layer perceptron, verify the conditional expectation of the influence of the abnormal scenario operation performance characteristics corresponding to the numerical values of the hardware nodes and software nodes in the causal structure diagram of the abnormal scenario operation of the known target server cluster components, and obtain the causal intervention characterization of the abnormal scenario association between the hardware and software of the known target server cluster components; Determine the standardized operation parameters of the hardware nodes - software nodes of the known target server cluster components, and form an array of standardized parameters of the known target server cluster components; Normalize the causal intervention characterization of the abnormal scenario association between the hardware and software of the known target server cluster components, and assign weights to the characteristic time series data of the abnormal scenario operation performance of the known target server cluster components; Using the Euclidean distance formula, with the weight of the abnormal scenario operation performance characteristics of the known target server cluster components as an external intervention factor, calculate the spatial distance between the characteristic time series data of the abnormal scenario operation performance of the known target server cluster components and the array of standardized parameters of the known target server cluster components, and determine the current abnormal deviation state index of the target server cluster components;

[0018] Among them, is the current abnormal deviation state index of the i-th component of the target server cluster, is the weight of the operation performance characteristics of the j-th abnormal scenario of the i-th component of the known target server cluster, is the operation performance characteristic value of the j-th abnormal scenario of the i-th component of the known target server cluster, is the normalization parameter of the i-th component of the target server cluster, and n is the total number of components of the target server cluster.

[0019] When in use, combined with the content in the above steps, As further content, the existing abnormal correlation analysis of server clusters mainly relies on simple statistics or rule engines, making it difficult to model complex causal chains; and DTW time series matching is only applicable to single-index analysis and cannot handle multi-component linkage; causal discovery techniques (such as Granger causality) are prone to misjudgment in heterogeneous hardware-software data; and root cause location methods (such as MicroRCA) rely on manual rules and cannot adapt to dynamic scenarios. Therefore, the existing solutions have defects such as static causal assumptions, lack of intervention effect analysis, distorted cross-modal distance calculation, and insufficient real-time performance.

[0020] This solution realizes accurate marking of cross-modal causal chains through joint modeling of DTW time series alignment and causal graphs, combines MLP intervention simulation to quantify the impact of parameter adjustment, uses normalized weights to strengthen key features, and verifies the deviation state of server cluster components under the operation performance characteristics of abnormal scenarios of components in the known target server cluster through a normalized parameter array and the Euclidean distance formula, significantly improving the accuracy of causal inference in complex abnormal scenarios.

[0021] Step 3: Based on the current abnormal deviation state of the components in the target server cluster, mark the abnormal component nodes and the corresponding associated task sequences, establish a resource allocation model for uninterrupted task migration, and generate repair tasks for abnormal scenarios of the server cluster; The said Step 3 includes the following content: Substitute the current associated task sequence to be executed by the components in the target server cluster into the full connection edge network of hardware nodes - software nodes of the server cluster, determine the component load resource demand characteristic parameters of the associated task sequence to be executed, and form the current associated task demand characteristic vector of the components in the target server cluster; Use the full connection edge network of hardware nodes - software nodes of the server cluster to verify the resource conflict intensity between the current associated task demand characteristic vectors of the components in the target server cluster per unit time, and determine the priority of the current associated task demand of the components in the target server cluster; Based on Logistic regression, with the current abnormal deviation state index of the components in the target server cluster as the given condition, the component load demand characteristic data of the associated task sequence to be executed per unit time as the output, and the stop risk probability of the components in the target server cluster executing the associated task as the label data output; Determine the initialization performance parameters of the target server cluster component according to the construction parameters of the target server cluster component, and establish a digital twin model of the target server cluster component. Update the digital twin model of the target server cluster component by using the shutdown risk probability of the associated tasks executed by the server cluster component, and determine the loadable performance support vector of the target server cluster component. Use the loadable performance support vector of the target server cluster component to train the SVR support vector machine. Taking the priority of the current associated tasks to be executed by the target server cluster component as a constraint condition and maximizing the completion of the associated tasks to be executed as the objective function, generate the hyperplane boundary of the loadable performance execution task sequence to be executed by the target server cluster component, with the demand feature vector of the current associated tasks to be executed by the target server cluster component as the output and the abnormal scenario repair task of the server cluster as the output.

[0022] When in use, combine the content in the above steps. As further content, the current task migration technology (such as the Kubernetes scheduler) mainly relies on static resource quotas and simple rules. Resource conflict detection is only based on single-dimensional threshold alarms. Digital twin applications are limited to physical device simulation and lack software task integration. The risk prediction model cannot handle complex anomalies. As a result, the rigid migration strategy cannot perceive hidden hardware anomalies, local optimization ignores the cascading effect of tasks, risk quantification lacks real-time anomaly association, and the digital twin model has a lag in update defects.

[0023] This solution realizes the unified representation of multi-dimensional constraints of hardware-software through demand feature vector modeling, and improves the accuracy of complex anomaly recognition by combining dynamic conflict assessment; uses Logistic regression and real-time digital twin to achieve high-precision risk prediction and intervention simulation; optimizes the task completion rate based on the SVR hyperplane boundary; ensures key task resources through priority-aware scheduling; and improves the accuracy of accurate prediction of task risks and optimal resource allocation.

[0024] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for generating a repair task for an abnormal scenario of a server cluster based on big data AI, characterized in that, Including: S1. Obtain the historical operation logs and real-time operation data of the server cluster, analyze the multimodal anomaly fusion of the interaction relationship between the hardware layer and the software layer of the server cluster, and generate the abnormal scenario behavior data of the target server cluster component operation; S2. Conduct a causal association analysis on the abnormal scenario behavior data of the target server cluster component operation and the big data of the known server cluster abnormal scenario behavior performance, and evaluate the current abnormal deviation state of the target server cluster component; S3. Based on the current abnormal deviation state of the target server cluster component, mark the abnormal component nodes and the corresponding associated task sequences, establish a non-interrupted task migration resource allocation model, and generate the server cluster abnormal scenario repair tasks.

2. The method for generating an abnormal scenario repair task for a server cluster based on big data AI according to claim 1, wherein The S1 includes: According to the unit timestamp, collect the original performance parameters of each component of the server cluster, perform preprocessing on the original performance parameters of each component using sliding window normalization, and form the hardware performance normalization vector of the server cluster; According to the unit timestamp, collect the historical operation logs of each component of the server cluster, train the BERT model, extract the log semantic set of the server cluster and substitute it into the Softmax function to assign weights to the log semantics of the server cluster; Using the attention weighting method, taking the abnormal scenario of the known server cluster component operation as the query matrix, the log semantic set as the query key, and the hardware performance normalization vector of the server cluster as the query value, calculate the log feature vector of the server cluster.

3. The method for generating an abnormal scenario repair task of a server cluster based on big data AI according to claim 2, wherein The S1 further includes: Taking the hardware performance normalization vector of the server cluster as the hardware node and the log feature vector of the server cluster as the software node, calculate the correlation coefficient between the hardware node and the software node using the Pearson correlation coefficient to generate the hardware node-software node connection edge, and use the Sigmoid to assign weights to the full connection edge of the hardware node-software node to establish the full connection edge network of the hardware node-software node of the server cluster; Extract each adjacent node in the full connection edge network of the hardware node-software node of the server cluster, establish the adjacent node array of the server cluster, and form the hardware node-software adjacent node matrix of the server cluster; Use the hardware node-software adjacent node matrix of the server cluster to train the GCN graph convolutional network, calculate the abnormal probability of each hardware node-software adjacent node using the full connection layer, and form the abnormal scenario behavior data of the target server cluster component operation.

4. The method for generating an abnormal scenario repair task of a server cluster based on big data AI according to claim 3, wherein, The S2 includes: Using the DTW dynamic event regularization algorithm, recursively calculate the maximum similarity between the big data of the known server cluster abnormal scenario behavior performance and the abnormal scenario behavior data of the target server cluster component operation, screen out the characteristic parameters of the known server cluster abnormal scenario behavior performance, and establish the characteristic time series data of the abnormal scenario operation performance of the known target server cluster component; Based on the characteristic time series data of the abnormal scenario operation performance of the known target server cluster component, substitute it into the full connection edge network of the hardware node-software node of the server cluster, and establish the causal graph of the abnormal scenario operation of the known target server cluster component according to the computing power call path of the hardware node-software node. Using the independent condition testing method, verify the correlation strength of each edge in the causal graph of the abnormal scenario operation of the known target server cluster components, determine the causal chain between each node, and obtain the causal structure diagram of the abnormal scenario operation of the known target server cluster components; Based on the MLP multi-layer perceptron, verify the influence condition expectation of the abnormal scenario operation performance characteristics corresponding to the numerical values of the hardware node and the software node in the causal structure diagram of the abnormal scenario operation of the known target server cluster components, and obtain the causal intervention characterization of the abnormal scenario of the known target server cluster components related to hardware-software.

5. The method for generating an abnormal scenario repair task for a server cluster based on big data AI according to claim 4, wherein The S2 further includes: Determine the hardware node-software node operation standardization parameters of the known target server cluster components, and form a standardization parameter array of the known target server cluster components; Normalize the causal intervention characterization of the abnormal scenario of the known target server cluster components related to hardware-software, and assign weights to the time series data of the abnormal scenario operation performance characteristics of the known target server cluster components; Using the Euclidean distance formula, with the weight of the abnormal scenario operation performance characteristics of the known target server cluster components as an external intervention factor, calculate the spatial distance between the time series data of the abnormal scenario operation performance characteristics of the known target server cluster components and the standardization parameter array of the known target server cluster components, and determine the current abnormal deviation state index of the target server cluster components; ； Among them, is the current abnormal deviation status index of the i-th component of the target server cluster, is the weight of the running performance characteristics of the j-th abnormal scenario of the i-th component of the known target server cluster, is the running performance characteristic value of the j-th abnormal scenario of the i-th component of the known target server cluster, is the normalization parameter of the i-th component of the target server cluster, and n is the total number of components of the target server cluster.

6. The method for generating an abnormal scenario repair task of a server cluster based on big data AI according to claim 5, wherein The S3 includes: Based on the current to-be-executed associated task sequence of the target server cluster components, substitute it into the hardware node-software node fully connected edge network of the server cluster, determine the component load resource requirement characteristic parameters of the to-be-executed associated task sequence, and form the current to-be-executed associated task requirement characteristic vector of the target server cluster components; Using the hardware node-software node fully connected edge network of the server cluster, verify the resource conflict strength between the current to-be-executed associated task requirement characteristic vectors of the target server cluster components per unit time, and determine the priority of the current to-be-executed associated task requirements of the target server cluster components.

7. The method for generating an abnormal scenario repair task of a server cluster based on big data AI according to claim 6, wherein The S3 further includes: Based on Logistic regression, with the current abnormal deviation state index of the target server cluster components as a given condition, the component load demand characteristic data per unit time of the to-be-executed associated task sequence as the output, and the shutdown risk probability of the target server cluster components executing the associated task as the labeled data output; According to the construction parameters of the target server cluster components, determine the initialization performance parameters of the target server cluster components, and establish a digital twin model of the target server cluster components, Update the digital twin model of the target server cluster components using the shutdown risk probability of the server cluster components executing the associated task, and determine the loadable performance support vector of the target server cluster components; Using the loadable performance support vector of the target server cluster components, train the SVR support vector machine. Taking the priority of the current associated tasks to be executed by the target server cluster components as the limiting condition and maximizing the completion of the associated tasks to be executed as the objective function, generate the hyperplane boundary of the loadable performance execution task sequence of the target server cluster components to be executed. Taking the feature vector of the current associated tasks to be executed by the target server cluster components as the output and the server cluster anomaly scenario repair task as the output.

Images