Data center intelligent operation and maintenance and fault prediction system and method

Abstract

The invention provides an intelligent operation and maintenance and fault prediction system and method for a data center, and the system comprises a self-healing decision module, a closed-loop verification module, an automatic execution module, a monitoring feedback module, and a strategy optimization module, and is characterized in that the self-healing decision module is used for determining the type and severity of a fault. The self-healing decision tree is generated according to the fault type and severity in combination with expert knowledge and historical data, the self-healing accuracy and safety are improved, the self-healing decision is executed through the automatic execution module, manual intervention is reduced, the operation and maintenance efficiency is improved, the self-healing process is monitored through the monitoring feedback module, feedback information is collected, and the operation and maintenance efficiency is improved. The self-healing effect is evaluated, the correctness and effectiveness of self-healing operation are ensured, a self-healing strategy is continuously optimized through a strategy optimization module according to feedback information and new data, the self-adaptive capacity and long-term performance of the system are improved, the system can continuously learn and adapt to a new fault mode, and therefore the self-healing capacity is continuously improved.

Description

Technical Field

[0001] The present invention relates to a fault prediction system and method, specifically a data center intelligent operation and maintenance and fault prediction system and method, belonging to the technical field of data center maintenance. Background Art

[0002] The data center intelligent operation and maintenance and fault prediction system is a comprehensive solution based on artificial intelligence (AI), machine learning (ML), the Internet of Things (IoT), and big data analysis technologies. It aims to improve data center operation and maintenance efficiency, reduce failure rates, and optimize resource utilization through automated and intelligent means. Its core goal is to detect potential problems in advance, reduce downtime, and ensure business continuity through real-time monitoring, data analysis, and predictive models. The system consists of two core modules: intelligent operation and maintenance and fault prediction. Intelligent operation and maintenance can automatically complete tasks such as configuration updates, log analysis, and anomaly detection, reducing manual intervention. Fault prediction can dynamically adjust alarm thresholds based on operating status to avoid false alarms and missed alarms.

[0003] In data center intelligent operations and fault prediction systems, self-healing capabilities are key to achieving efficient and reliable operations. However, current systems have significant limitations in self-healing capabilities, primarily due to three core issues: complexity of decision logic, execution security limitations, and a single self-healing strategy. Complexity in decision logic: Data center failures are often caused by a combination of factors (such as network congestion, hardware aging, and software configuration errors), requiring comprehensive analysis of multiple data sources such as logs, metrics, and topology. Existing systems often rely on static rules or simple models (such as threshold judgments) and have difficulty handling complex relationships. Furthermore, data center environments (such as traffic patterns and load distribution) change over time, making fixed rules difficult to adapt. Execution security restrictions: Automatic repair operations often lead to new problems. For example, incorrectly isolating a critical service may cause the business modules that rely on it to crash. Certain operations cannot be rolled back, and execution errors may cause permanent losses. The existing system permissions are too coarse-grained to assign refined permissions for specific fault types. Single self-healing strategy: Existing systems are mostly based on fixed rules (such as "add capacity if CPU overloaded"), making them difficult to handle new types of failures (such as zero-day attacks and hidden hardware failures). As data centers scale, the number of rules increases exponentially, making manual maintenance extremely difficult. Furthermore, these systems lack contextual awareness and fail to consider the context of the failure (such as time and business priority), leading to irrational decision-making. Furthermore, self-healing strategies rely on manual updates for a long time and cannot adapt to changes in failure modes in real time. To this end, a data center intelligent operation and maintenance and fault prediction system and method are proposed. Summary of the Invention

[0004] In view of this, the present invention provides a data center intelligent operation and maintenance and fault prediction system and method to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.

[0005] The technical solution of the embodiment of the present invention is implemented as follows: a data center intelligent operation and maintenance and fault prediction system, including a self-healing decision module, a closed-loop verification module, an automated execution module, a monitoring feedback module, and a strategy optimization module: The self-healing decision module is used to formulate a self-healing strategy according to the fault type and severity; The self-healing decision module uses a decision algorithm to formulate strategies, including fault classification, severity grading, decision tree generation and self-healing strategy selection; The self-healing strategies include resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing and business self-healing; The closed-loop verification module is used to verify the immediate effect of the self-healing strategy using a verification algorithm, including data collection, effect evaluation and strategy adjustment; The automated execution module is used to convert the self-healing strategy formulated by the self-healing decision module into specific operation instructions and execute them; The monitoring feedback module is used to collect system status data in real time, including CPU usage, memory usage, network latency, and service availability, and identify abnormal indicators based on threshold rules to trigger the self-healing process. During the self-healing process, the monitoring feedback module records the operation instructions and execution time generated by the self-healing decision module, and maintains heartbeat communication with the automation execution module to confirm whether the operation is successfully delivered to the target node. The strategy optimization module is used to achieve dynamic iterative updates of the self-healing strategy through data-driven and feedback closed-loop mechanisms, including four stages: data collection, model retraining, strategy evaluation and selection, and deployment verification.

[0006] Further preferably, the decision algorithm includes: Fault classification: The acquired fault types include hardware failure, network congestion, and software anomalies, and the severity of the fault types is scored based on the impact scope, duration, and business priority. Severity rating: The severity is categorized into low, medium, high, and urgent. Low severity refers to single-node performance degradation that does not affect business continuity, while urgent severity refers to core service unavailability, resulting in business interruption. Generate a decision tree: Based on the expert knowledge base and historical failure cases, an initial decision tree is constructed. The decision tree branches are dynamically pruned based on the current load, resource utilization, and business priorities. Self-healing strategy selection: A weighted scoring model is used to comprehensively consider recovery time, data security, and operational risk factors to select a self-healing strategy.

[0007] Further preferably, the verification algorithm comprises the following steps: Multi-dimensional data collection: Real-time acquisition of key indicators after self-healing operations, including business-level metrics such as request success rate, response time, and throughput; system-level metrics such as CPU / memory utilization, disk IOPS, and network latency; and application-level metrics such as number of database connections, cache hit rate, and thread pool status. Application logs and self-healing operation logs are analyzed to extract operation time, execution action, and resource change information. User reporting interfaces are integrated to collect data on service quality perceptions from business parties. Effect evaluation model: Based on the system status before the self-healing operation, recovery thresholds are set and evaluation indicator weights are dynamically adjusted based on business priorities. Evaluation dimensions include: Validity: whether the fault is eliminated; Efficiency: whether the recovery time meets the requirements; Side effects: whether secondary failures are caused; Resource cost: Whether the resources consumed by the self-healing operation exceed the budget; Strategy adjustment: Store self-healing cases in the knowledge graph to enrich decision-making basis, convert evaluation results into reward signals, and trigger the strategy center to optimize the decision tree.

[0008] Further preferably, the strategy optimization module implements dynamic iterative update of the self-healing strategy, comprising the following steps: Data Collection: Obtain self-healing operation logs, including operation type, execution time, and effect evaluation; real-time fault data, including new attack signatures and hidden hardware failures; and environmental context, including integrated business priorities, maintenance windows, and resource utilization. Strategy evaluation: Use historical data to simulate the effects of new strategies, calculate metrics including success rate and resource consumption, conduct online A / B testing, run new and old strategies in parallel in a small environment, and compare the effects in real time; Deployment verification: Gradually deploy the new policy to the production environment, monitor key indicators, including self-healing success rate and error rate, and establish a rollback mechanism. If the new policy causes performance degradation, automatically roll back to the old version and trigger an alarm.

[0009] Further preferably, it also includes a multi-source heterogeneous data acquisition module, a feature extraction module, a fault pattern recognition module and a fault warning module; The multi-source heterogeneous data acquisition module is used to collect raw data from devices, including servers, network devices, storage devices, databases, sensors, cameras, and cloud platforms. The raw data includes log data, performance indicators, environmental parameters, business data, configuration data, and security data.

[0010] Further preferably, the feature extraction module is used to extract specific features from the data, and the features include statistical features, spectral features, time domain features, machine learning features and business-specific features.

[0011] Further preferably, the fault pattern recognition module is used to identify the fault type using a fault recognition algorithm, and the fault recognition algorithm includes the following steps: Data collection: Collect various data related to the fault, including sensor data, log data, and performance indicators; Feature extraction: Extract features related to fault identification from preprocessed data, including statistical features, spectral features, time domain features, and machine learning features; Algorithm selection and training: Select a neural network algorithm for fault identification, use the extracted features as input, and train and optimize the algorithm; Model evaluation and verification: Build a fault identification model, use a test dataset to evaluate and verify the trained model, and adjust and optimize the model based on the evaluation results; Deployment and application: Deploy the trained model to the actual application environment for real-time monitoring and fault identification.

[0012] Further preferably, the fault types include damage type fault, degradation type fault, loose type fault, imbalance type fault, blockage and leakage type fault, performance degradation or functional failure type fault.

[0013] Further preferably, the fault warning module is used to determine whether a warning needs to be triggered based on preset thresholds and rules, and to notify staff via email or text message.

[0014] A data center intelligent operation and maintenance and fault prediction method includes the following steps: Step 1: Collect raw data from the device and extract specific features from the data, including statistical features, spectral features, time domain features, machine learning features, and business-specific features; Step 2: Use fault identification algorithms to identify fault types, including damage faults, degradation faults, loose faults, imbalance faults, blockage and leakage faults, performance degradation or functional failure faults; Step 3: Determine whether an alert needs to be triggered based on preset thresholds and rules, and notify staff via email or SMS; Step 4: Develop a self-healing strategy based on the fault type and severity, including resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing, and business self-healing. Use a verification algorithm to verify the immediate effectiveness of the self-healing strategy. Step 5: Convert the self-healing strategy into specific operational instructions and execute them, collecting system status data in real time; Step 6: Realize dynamic iterative updates of self-healing strategies through data-driven and feedback closed-loop mechanisms.

[0015] The embodiment of the present invention adopts the above technical solution, which has the following advantages: The present invention generates a self-healing decision tree based on the fault type and severity, combined with expert knowledge and historical data, and selects the best self-healing solution, thereby improving the accuracy and safety of self-healing. The automatic execution module executes self-healing decisions, such as restarting services and adjusting resources, which reduces manual intervention and improves operation and maintenance efficiency. The monitoring feedback module monitors the self-healing process, collects feedback information, evaluates the self-healing effect, and ensures the correctness and effectiveness of the self-healing operation. The strategy optimization module continuously optimizes the self-healing strategy based on feedback information and new data, thereby improving the system's adaptability and long-term performance, enabling the system to continuously learn and adapt to new fault modes, thereby continuously improving its self-healing capability.

[0016] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 This is a structural diagram of a data center intelligent operation and maintenance and fault prediction system of the present invention; Figure 2 is a flowchart of the steps of the decision-making algorithm of the present invention; Figure 3 A flowchart of the steps for dynamic iterative update of the self-healing strategy of the present invention; Figure 4 is a flowchart of the steps of the fault identification algorithm of the present invention; Figure 5 This is a flowchart of the steps of a data center intelligent operation and maintenance and fault prediction method of the present invention. DETAILED DESCRIPTION

[0019] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0020] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] like Figure 1-Figure 5 As shown, an embodiment of the present invention provides a data center intelligent operation and maintenance and fault prediction system, including a self-healing decision module, a closed-loop verification module, an automated execution module, a monitoring feedback module, and a strategy optimization module; The self-healing decision module is used to formulate self-healing strategies based on the fault type and severity; The self-healing decision module uses a decision-making algorithm to formulate strategies, including fault classification, severity grading, decision tree generation, and self-healing strategy selection; The decision-making algorithm includes: Fault classification: The acquired fault types include hardware failure, network congestion, and software anomalies, and the severity of the fault types is scored based on the impact scope, duration, and business priority. Severity rating: The severity is categorized into low, medium, high, and urgent. Low severity refers to single-node performance degradation that does not affect business continuity, while urgent severity refers to core service unavailability, resulting in business interruption. Generate a decision tree: Based on the expert knowledge base and historical failure cases, an initial decision tree is constructed. The decision tree branches are dynamically pruned based on the current load, resource utilization, and business priorities. Self-healing strategy selection: A weighted scoring model is used to comprehensively consider recovery time, data security, and operational risk factors to select a self-healing strategy. The logic of the decision algorithm is: Input: fault type (F), severity (S), context information (C, such as time, business priority); Output: Self-healing plan (A); step: Root node judgment: Determine the initial branch based on the fault type (F) and severity (S) (e.g., "hardware fault - high severity" enters branch A1).

[0022] Context-sensitive branching: Context information (C) is used to further refine the solution (e.g., low-priority services at night can choose "restart service" and high-priority services during the day can choose "switch to disaster recovery environment"). Risk assessment: Monte Carlo simulation is used to assess operational risks. If the risk exceeds the threshold, a manual review is triggered. Self-healing strategies include resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing, and business self-healing. The adjustment actions caused by resource reconstruction include virtual machine hot migration and container scheduling; Adjustment actions generated by service self-healing include automatic restart, configuration rollback, and rate limiting and degradation; The adjustment actions generated by network self-healing include link switching, load balancing adjustment, and QoS reconfiguration; The adjustment actions generated by data self-recovery include master-slave switching, replica reconstruction, and data verification; Adjustments made by hardware self-healing include power switching, fan speed adjustment, and hard drive bad sector isolation; Adjustments resulting from service self-recovery include switching traffic to the backup site and activating the disaster recovery unit; The closed-loop verification module is used to verify the immediate effect of the self-healing strategy using a verification algorithm, including data collection, effect evaluation, and strategy adjustment; The verification algorithm consists of the following steps: Multi-dimensional data collection: Real-time acquisition of key indicators after self-healing operations, including: Business layer: request success rate, response time, and throughput; System layer: CPU / memory utilization, disk IOPS, network latency; Application layer: number of database connections, cache hit rate, thread pool status; Parse application logs and self-healing operation logs to extract operation time, execution action, and resource change information, integrate user reporting interfaces, and collect service quality perception data from business parties. Effect evaluation model: Based on the system status before the self-healing operation, recovery thresholds are set and evaluation indicator weights are dynamically adjusted based on business priorities. Evaluation dimensions include: Validity: whether the fault is eliminated, such as whether the error log stops being generated; Efficiency: Whether the recovery time meets the requirements, for example, a database failure must be recovered within 3 minutes; Side effects: whether secondary failures are caused, such as whether other nodes are overloaded after load balancing adjustments; Resource cost: Whether the resources consumed by the self-healing operation exceed the budget, such as whether the VM migration causes excessive host load; Strategy Adjustment: Self-healing cases are stored in the knowledge graph to enrich decision-making evidence. Evaluation results are converted into reward signals, triggering the strategy center to optimize the decision tree. For example, if a self-healing strategy causes network congestion three times in a row, its probability of being selected is reduced. If a new strategy performs better than the existing strategy in a simulation environment, its weight is gradually increased. The verification algorithm also includes manual intervention and exception handling: Expert review: Cases with questionable assessment results (e.g., indicator fluctuations with no clear correlation to self-healing operations) are sent to the expert system for review; Circuit breaker mechanism: When the self-healing operation failure rate exceeds the preset threshold (e.g., five consecutive unresolved failures), the strategy is automatically suspended and a manual alarm is triggered; Root cause tracing: Combined with the digital twin module, the fault scenario is reproduced in a virtual environment to verify the correctness of the self-healing logic; Apply the verification algorithm to real-world scenarios: For example, if a database node triggers self-healing due to a disk error, the system automatically migrates the instance to a standby node; Verification process: Collect new node performance indicators within 30 minutes after migration; Compare the business response time (must be ≤200ms) and TPS (must be ≥800) before and after the migration. Check whether the load of the associated storage array is balanced; If the network delay of the new node increases suddenly (for example, exceeding the threshold of 5ms), the self-healing is considered to have failed, triggering a rollback and an alarm. Results: This strategy reduced the MTTR of database failures from 25 minutes to 8 minutes without causing any secondary problems.

[0023] The automated execution module is used to convert the self-healing strategy formulated by the self-healing decision module into specific operation instructions and execute them; The steps to achieve automation include: Operation atomicization: Decompose complex self-healing processes into uninterruptible atomic operations, such as restarting a single service; Dependency management: identifying dependencies between operations and ensuring they are executed in order; Status tracking: real-time monitoring of operation execution status to ensure successful completion of each step; Rollback mechanism: If an operation fails, it can automatically roll back to the previous state; When applied to actual scenarios: For example, a securities trading platform detects a memory leak in the core trading server before the market opens; Execution process: The decision engine outputs a strategy: complete faulty server isolation and business migration one hour before market opening. The orchestration engine automatically: Allocate a virtual machine with the same configuration from the spare pool; Perform hot migration (memory dirty page rate is controlled to <3%); Verify transaction interface response time (must be <50ms); Update F5 load balancer configuration; Result: The business switch took 82 seconds, the system load was stable at the opening, and there was no transaction loss.

[0024] The monitoring feedback module collects system status data in real time, including CPU usage, memory usage, network latency, and service availability. It identifies abnormal indicators based on threshold rules and triggers the self-healing process. During the self-healing process, it records the operation instructions and execution time generated by the self-healing decision module and maintains heartbeat communication with the automated execution module to confirm whether the operation has been successfully delivered to the target node. The specific workflow of the monitoring feedback module includes: 1. Real-time monitoring stage Data collection: Collect system status data (such as CPU usage, memory usage, network latency, and service availability) in real time through agents or probes; Anomaly detection: Identify abnormal indicators based on threshold rules or machine learning models (such as Isolation Forest) and trigger self-healing processes; Example: During a promotional event in an e-commerce system, the monitoring module detects a sudden increase in the number of database connections exceeding a threshold (80% → 95%) and automatically marks it as an anomaly. 2. Self-healing execution tracking phase Operation log records: record the operation instructions (such as "expand instance" and "restart service") generated by the self-healing decision module and the execution time; State synchronization: maintains heartbeat communication with the automation execution module to confirm whether the operation is successfully sent to the target node; Example: The self-healing system decides to "expand the database read replica." The monitoring module records the start time of the operation (14:00) and tracks the instance status (from "Creating" to "Running"). 3. Effect evaluation stage Indicator comparison: Compare key indicators (such as response time and error rate) before and after self-healing and calculate the improvement; Root cause verification: Combined with the output of the fault pattern identification module, confirm whether the self-healing operation solves the root cause of the problem; Example: After capacity expansion, the number of database connections dropped to 70%, and the query response time dropped from 2s to 500ms, verifying that the capacity expansion was effective. 4. Feedback generation stage Report generation: Automatically generate a self-healing effect report, including operation details, indicator changes, and root cause analysis results; Risk warning: If the self-healing operation does not meet expectations (for example, the latency remains high after capacity expansion), it is marked as a high-risk event and triggers manual intervention; Example: The report states, "CPU usage remains above 90% after capacity expansion. Further inspection of slow query logs is recommended." 5. Closed-loop optimization stage Policy iteration: Feedback the evaluation results to the self-healing decision module to adjust weight parameters or rule thresholds; Knowledge base update: Successful self-healing cases are added to the knowledge graph, and failed cases are recorded in the question database for subsequent analysis; Example: Based on multiple capacity expansion cases, optimize the triggering conditions for the "Database Performance Issue → Capacity Expansion" branch in the decision tree.

[0025] The policy optimization module is used to achieve dynamic iterative updates of self-healing strategies through data-driven and feedback closed-loop mechanisms. It includes four stages: data collection, model retraining, policy evaluation and selection, and deployment verification. The policy optimization module implements dynamic iterative updates of the self-healing policy, including the following steps: Data Collection: Obtain self-healing operation logs, including operation type, execution time, and effect evaluation; real-time fault data, including new attack signatures and hidden hardware failures; and environmental context, including integrated business priorities, maintenance windows, and resource utilization. Strategy evaluation: Use historical data to simulate the effects of new strategies, calculate metrics including success rate and resource consumption, conduct online A / B testing, run new and old strategies in parallel in a small environment, and compare the effects in real time; Deployment verification: Gradually deploy the new policy to the production environment, monitor key indicators, including self-healing success rate and error rate, and establish a rollback mechanism. If the new policy causes performance degradation, automatically roll back to the old version and trigger an alarm.

[0026] In one embodiment, it also includes a multi-source heterogeneous data acquisition module, a feature extraction module, a fault pattern recognition module and a fault early warning module; The multi-source heterogeneous data acquisition module is used to collect raw data from devices, including servers, network devices, storage devices, databases, sensors, cameras, and cloud platforms. Wherein servers include physical servers, virtual servers and cloud instances; Network equipment includes routers, switches, firewalls, and load balancers; Storage devices include disk arrays, network-attached storage, and storage area networks; Databases include relational databases and NoSQL databases; Raw data includes log data, performance indicators, environmental parameters, business data, configuration data, and security data; The log data includes: System log: operating system events (such as login, file access), service start / stop records; Application logs: business operation records (such as order creation, successful payment), error stack information; Security logs: Firewall rule matches, unusual login attempts Performance indicators include: CPU: usage, load average, context switches; Memory: used / free memory, swap partition usage; Disk I / O: read and write rates, IOPS (input and output operations per second); Network: bandwidth utilization, packet loss rate, and latency; Environmental parameters include: Physical environment: temperature, humidity, power consumption (data center PUE value); Virtual environment: container resource quotas (CPU / memory limits), virtual machine snapshot status; Business data includes: Transaction data: order amount, payment status, user behavior path; User data: registration information, login frequency, preferences; Configuration data includes: Device configuration: network device routing tables, server BIOS settings; Application configuration: database connection pool size, cache expiration time; Safety data includes: Threat intelligence: malicious IP addresses, vulnerability scan results; Compliance data: data access audit logs, encryption status.

[0027] The multi-source heterogeneous data acquisition module collects multi-dimensional data such as logs, performance, environment, business, configuration, and security by covering a wide range of sources such as IT infrastructure, application systems, terminal devices, cloud platforms, and security systems, providing a comprehensive and real-time information foundation for the self-healing system. In actual applications, appropriate collection tools and technologies must be selected according to specific scenarios to ensure data accuracy and availability.

[0028] In one embodiment, the feature extraction module is used to extract specific features from the data, including statistical features, spectral features, time domain features, machine learning features, and business-specific features; Statistical features include: Average: Calculate the average value of the data series to reflect the central trend of the data; Standard deviation: measures the dispersion of a data series, that is, how dispersed the data points are around the mean; Maximum / minimum values: Identify extreme values in a data series, which helps to detect anomalies or extreme situations; Median: When the data is unevenly distributed or there are outliers, the median represents the center of the data better than the mean; Spectral characteristics include: Frequency component: Through Fourier transform, the data is converted from the time domain to the frequency domain to analyze the frequency components in the data; Power spectral density: describes the power distribution of a signal at different frequencies, which helps identify periodic components or noise in the signal; Time domain features include: Autocorrelation function: measures the correlation of a signal with itself at different time delays, which helps identify periodic or repetitive patterns in the signal; Waveform characteristics: such as rise time, fall time, peak value, etc., describe the shape and changes of the signal waveform; Machine learning features include: Principal Component Analysis (PCA) features: Through dimensionality reduction technology, high-dimensional data is converted into low-dimensional features, retaining the main information of the data; Clustering features: clustering algorithms are used to group data points and extract representative features of each group of data; Deep learning features: using deep neural networks to automatically learn complex features and patterns in data; Business-specific characteristics include: Based on specific business scenarios and needs, extract features closely related to the business. For example, in e-commerce systems, extract features such as order amount, user purchase frequency, and product category distribution; in IoT systems, extract sensor data features such as device temperature, humidity, and pressure.

[0029] In one embodiment, the fault pattern recognition module is used to identify the fault type using a fault recognition algorithm, the fault recognition algorithm comprising the following steps: Data collection: Collect various data related to the fault, including sensor data, log data, and performance indicators; Feature extraction: Extract features related to fault identification from preprocessed data, including statistical features, spectral features, time domain features, and machine learning features; Algorithm selection and training: Select a neural network algorithm for fault identification, use the extracted features as input, and train and optimize the algorithm; Model evaluation and verification: Build a fault identification model, use a test dataset to evaluate and verify the trained model, and adjust and optimize the model based on the evaluation results; Deployment and application: Deploy the trained model to the actual application environment for real-time monitoring and fault identification.

[0030] In one embodiment, the fault types include damage faults, degradation faults, loose faults, imbalance faults, blockage and leakage faults, performance degradation or functional failure faults; the fault warning module is used to determine whether a warning needs to be triggered based on preset thresholds and rules, and notify the staff via email or text message.

[0031] A data center intelligent operation and maintenance and fault prediction method includes the following steps: Step 1: Collect raw data from the device and extract specific features from the data, including statistical features, spectral features, time domain features, machine learning features, and business-specific features; Step 2: Use fault identification algorithms to identify fault types, including damage faults, degradation faults, loose faults, imbalance faults, blockage and leakage faults, performance degradation or functional failure faults; Step 3: Determine whether an alert needs to be triggered based on preset thresholds and rules, and notify staff via email or SMS; Step 4: Develop a self-healing strategy based on the fault type and severity, including resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing, and business self-healing. Use a verification algorithm to verify the immediate effectiveness of the self-healing strategy. Step 5: Convert the self-healing strategy into specific operational instructions and execute them, collecting system status data in real time; Step 6: Realize dynamic iterative updates of self-healing strategies through data-driven and feedback closed-loop mechanisms.

[0032] When the present invention is working: the multi-source heterogeneous data acquisition module collects raw data from the equipment and extracts specific features from the data; the fault pattern recognition module uses the fault recognition algorithm to identify the fault type, including damage type fault, degradation type fault, loose type fault, imbalance type fault, blockage and leakage type fault, performance degradation or functional failure type fault; the fault warning module determines whether it is necessary to trigger the warning according to the preset threshold and rules; the self-healing decision module formulates the self-healing strategy according to the fault type and severity; the closed-loop verification module uses the verification algorithm to verify the immediate effect of self-healing; the automatic execution module converts the self-healing strategy into specific operation instructions and executes them, and collects system status data in real time; the strategy optimization module realizes the dynamic iterative update of the self-healing strategy through data-driven and feedback closed-loop mechanism.

[0033] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various modifications and substitutions within the technical scope disclosed in the present invention, and such modifications and substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A data center intelligent operation and maintenance and fault prediction system, comprising a self-healing decision module, a closed-loop verification module, an automated execution module, a monitoring feedback module, and a strategy optimization module, characterized by: The self-healing decision module is used to formulate a self-healing strategy according to the fault type and severity; The self-healing decision module adopts a decision algorithm to formulate strategies, including fault classification, severity grading, decision tree generation and self-healing strategy selection; The self-healing strategies include resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing and business self-healing; The closed-loop verification module is used to verify the immediate effect of the self-healing strategy using a verification algorithm, including data collection, effect evaluation and strategy adjustment; The automated execution module is used to convert the self-healing strategy formulated by the self-healing decision module into specific operation instructions and execute them; The monitoring feedback module is used to collect system status data in real time, including CPU usage, memory usage, network latency, and service availability, and identify abnormal indicators based on threshold rules to trigger the self-healing process. During the self-healing process, the monitoring feedback module records the operation instructions and execution time generated by the self-healing decision module, and maintains heartbeat communication with the automation execution module to confirm whether the operation is successfully delivered to the target node. The strategy optimization module is used to achieve dynamic iterative updates of the self-healing strategy through data-driven and feedback closed-loop mechanisms, including four stages: data collection, model retraining, strategy evaluation and selection, and deployment verification.

2. The data center intelligent operation and maintenance and fault prediction system according to claim 1, characterized in that: The decision algorithm includes: Fault classification: The acquired fault types include hardware failures, network congestion, and software anomalies, and the severity of the fault types is scored based on the impact scope, duration, and business priority. Severity rating: The severity is categorized into low, medium, high, and urgent. Low severity refers to single-node performance degradation that does not affect business continuity, while urgent severity refers to core service unavailability, resulting in business interruption. Generate a decision tree: Based on the expert knowledge base and historical failure cases, an initial decision tree is constructed. The decision tree branches are dynamically pruned based on the current load, resource utilization, and business priorities. Self-healing strategy selection: A weighted scoring model is used to comprehensively consider recovery time, data security, and operational risk factors to select a self-healing strategy.

3. The data center intelligent operation and maintenance and fault prediction system according to claim 1, characterized in that: The verification algorithm comprises the following steps: Multi-dimensional data collection: Real-time acquisition of key indicators after self-healing operations, including business-level indicators such as request success rate, response time, and throughput; System layer: CPU / memory utilization, disk IOPS, and network latency; Application layer: Database connection count, cache hit rate, and thread pool status; parse application logs and self-healing operation logs to extract operation time, execution action, and resource change information; integrate user reporting interfaces to collect service quality perception data from business parties; Effect evaluation model: Based on the system status before the self-healing operation, recovery thresholds are set and evaluation indicator weights are dynamically adjusted based on business priorities. Evaluation dimensions include: Validity: whether the fault is eliminated; Efficiency: whether the recovery time meets the requirements; Side effects: whether secondary failures are caused; Resource cost: Whether the resources consumed by the self-healing operation exceed the budget; Strategy adjustment: Store self-healing cases in the knowledge graph to enrich decision-making basis, convert evaluation results into reward signals, and trigger the strategy center to optimize the decision tree.

4. The data center intelligent operation and maintenance and fault prediction system according to claim 1, characterized in that: The strategy optimization module implements dynamic iterative updates of the self-healing strategy, including the following steps: Data Collection: Obtain self-healing operation logs, including operation type, execution time, and effect evaluation; real-time fault data, including new attack signatures and hidden hardware failures; and environmental context, including integrated business priorities, maintenance windows, and resource utilization. Strategy evaluation: Use historical data to simulate the effects of new strategies, calculate metrics including success rate and resource consumption, conduct online A / B testing, run new and old strategies in parallel in a small environment, and compare the effects in real time; Deployment verification: Gradually deploy the new policy to the production environment, monitor key indicators, including self-healing success rate and error rate, and establish a rollback mechanism. If the new policy causes performance degradation, automatically roll back to the old version and trigger an alarm.

5. The data center intelligent operation and maintenance and fault prediction system according to claim 1, characterized in that: It also includes a multi-source heterogeneous data acquisition module, a feature extraction module, a fault pattern recognition module and a fault warning module; The multi-source heterogeneous data acquisition module is used to collect raw data from devices, including servers, network devices, storage devices, databases, sensors, cameras and cloud platforms. The raw data includes log data, performance indicators, environmental parameters, business data, configuration data and security data.

6. The data center intelligent operation and maintenance and fault prediction system according to claim 5, characterized in that: The feature extraction module is used to extract specific features from the data, and the features include statistical features, spectral features, time domain features, machine learning features and business-specific features.

7. The data center intelligent operation and maintenance and fault prediction system according to claim 5, characterized in that: The fault pattern recognition module is used to identify the fault type using a fault recognition algorithm, wherein the fault recognition algorithm includes the following steps: Data collection: Collect various data related to the fault, including sensor data, log data, and performance indicators; Feature extraction: Extract features related to fault identification from preprocessed data, including statistical features, spectral features, time domain features, and machine learning features; Algorithm selection and training: Select a neural network algorithm for fault identification, use the extracted features as input, and train and optimize the algorithm; Model evaluation and verification: Build a fault identification model, use a test dataset to evaluate and verify the trained model, and adjust and optimize the model based on the evaluation results; Deployment and application: Deploy the trained model to the actual application environment for real-time monitoring and fault identification.

8. The data center intelligent operation and maintenance and fault prediction system according to claim 7, characterized in that: The fault types include damage type fault, degradation type fault, loose type fault, imbalance type fault, blockage and leakage type fault, performance degradation or functional failure type fault.

9. The data center intelligent operation and maintenance and fault prediction system according to claim 5, characterized in that: The fault warning module is used to determine whether a warning needs to be triggered based on preset thresholds and rules, and to notify staff via email or text message.

10. A data center intelligent operation and maintenance and fault prediction method, applied to a data center intelligent operation and maintenance and fault prediction system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: Collect raw data from the device and extract specific features from the data, including statistical features, spectral features, time domain features, machine learning features, and business-specific features; Step 2: Use fault identification algorithms to identify fault types, including damage faults, degradation faults, loose faults, imbalance faults, blockage and leakage faults, performance degradation and functional failure faults; Step 3: Determine whether an alert needs to be triggered based on preset thresholds and rules, and notify staff via email or SMS; Step 4: Develop a self-healing strategy based on the fault type and severity, including resource reconstruction, service self-healing, network self-healing, data self-healing, hardware self-healing, and business self-healing. Use a verification algorithm to verify the immediate effectiveness of self-healing. Step 5: Convert the self-healing strategy into specific operational instructions and execute them, collecting system status data in real time; Step 6: Realize dynamic iterative updates of self-healing strategies through data-driven and feedback closed-loop mechanisms.

Images