Operation and maintenance auxiliary method and system suitable for human and social department informatization system

Abstract

The invention provides an operation and maintenance auxiliary method and system suitable for a human and social department informatization system, and relates to the technical field of informatization system operation and maintenance, and the method comprises the steps: obtaining human and social department informatization system operation and maintenance data collected in real time, and carrying out the preprocessing; inputting operation and maintenance data into the trained model, and predicting future fault occurrence and resource requirements for assisting fault processing and resource scheduling of the system; wherein the operation and maintenance data is acquired in real time by adopting a mode of combining timing polling, event triggering and stream processing, and comprises virtual machine performance data, network state data, storage data, micro-service log data, AP I calling data, user behavior log data, safety monitoring data and performance monitoring data; based on artificial intelligence and operation and maintenance data, faults and resource use conditions are predicted, the fault positioning speed and efficient resource management are improved, the operation and maintenance cost is effectively reduced, and a solid technical support is constructed for improving service efficiency and user experience.

Description

Technical Field

[0001] The present invention belongs to the technical field of information system operation and maintenance, and in particular relates to an operation and maintenance auxiliary method and system suitable for an information system of a human resources and social security department. Background Art

[0002] The LEAF6 standard framework of human resources and social security, as the industry technical standard framework for the construction of application systems supporting the human resources and social security industry, fully adopts the microservice architecture; currently, various regions have successively established their own government cloud platforms, including government clouds that provide PAAS capabilities and government clouds that provide IAAS capabilities.

[0003] Most government clouds that provide PAAS capabilities are built on the three major cloud platform vendors Huawei, Alibaba, and Tencent, and provide rich and mature distributed components, such as distributed caching, distributed object storage, distributed messaging, microservice frameworks, etc. For users, they are ready to use and very friendly.

[0004] The government cloud that provides IAAS capabilities mainly provides virtual machines, and the corresponding distributed components need to be deployed and maintained by the application system developers themselves. In this case, there is a lack of monitoring and management of the PAAS layer cloud platform. Under the microservice architecture model, the labor cost of operation and maintenance increases sharply, and the operation and maintenance effect is not good. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides an operation and maintenance assistance method and system suitable for the information system of the human resources and social security department. Based on artificial intelligence and operation and maintenance data, faults and resource usage are predicted, the fault location speed and efficient resource management are improved, and the operation and maintenance costs are effectively reduced, so as to build a solid technical support for improving service efficiency and user experience.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention provides an operation and maintenance assistance method applicable to an information system of a human resources and social security department.

[0008] An operation and maintenance auxiliary method applicable to the information system of human resources and social security departments, comprising:

[0009] Obtain real-time collected operation and maintenance data of the human resources and social security department's information system and perform pre-processing;

[0010] The pre-processed operation and maintenance data is input into the trained model to predict future fault occurrence and resource requirements, which is used to assist in system fault handling and resource scheduling.

[0011] Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

[0012] Furthermore, the preprocessing includes removing duplicate, erroneous, and irrelevant data, removing noise, standardizing, and extracting features.

[0013] Furthermore, the feature extraction is to extract the features that characterize the status of the operating system, database, middleware and application system respectively, specifically:

[0014] Use file system usage, system and network log alarm information, CPU load, storage I / O performance, date, and time to represent the operating system status;

[0015] The database status is represented by database health status, ASM disk group and disk information, data file storage space information, table space usage, disaster recovery operation information, backup and archive log storage information, and monitoring logs.

[0016] Use node instance running status, log storage space information, and thread blocking information to represent the middleware status;

[0017] The application system status is represented by access information, system operation status information, database user memory usage, process waiting information, database lock information, microservice logs, API call data, and user behavior logs.

[0018] Furthermore, the model includes a fault prediction model and a resource prediction model;

[0019] The fault prediction model uses the time series of operation and maintenance data as input to predict the probability of faults and the type of faults;

[0020] The resource prediction model uses operation and maintenance data as input and is used to predict future file system usage, future CPU usage, and future memory usage.

[0021] Furthermore, the fault prediction model adds a softmax layer to perform classification on the basis of RNN and outputs the predicted probability of each fault type.

[0022] Furthermore, the resource prediction model uses a gradient boosting tree algorithm to construct a regression model to predict the file system usage, CPU usage, and memory usage at future time points.

[0023] Furthermore, the model also includes using an unsupervised learning model to detect abnormal data.

[0024] A second aspect of the present invention provides an operation and maintenance assistance system suitable for an information system of a human resources and social security department.

[0025] An operation and maintenance auxiliary system suitable for the information system of the human resources and social security department, including an acquisition module and a prediction module:

[0026] The acquisition module is configured to: acquire the real-time collected operation and maintenance data of the information system of the human resources and social security department and perform pre-processing;

[0027] The prediction module is configured to: input the pre-processed operation and maintenance data into the trained model to predict future fault occurrence and resource requirements to assist in system fault handling and resource scheduling;

[0028] Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

[0029] The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of an operation and maintenance assistance method applicable to an information system of a human resources and social security department as described in the first aspect of the present invention.

[0030] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of an operation and maintenance assistance method applicable to an information system of a human resources and social security department as described in the first aspect of the present invention are implemented.

[0031] One or more of the above technical solutions have the following beneficial effects:

[0032] 1. Intelligent Fault and Resource Prediction

[0033] By comprehensively collecting multi-dimensional data from different sources (including IaaS government cloud infrastructure, human resources and social security LEAF6 framework applications, and third-party monitoring systems), and applying it to the fault prediction model and resource prediction model constructed by machine learning algorithms, an automatic learning mode of historical fault data and resource growth trend characteristic data is realized, and the prediction of future faults and resource demands is realized; the model performance is evaluated through cross-validation, A / B testing and other methods to ensure the accuracy and reliability of the prediction results; this intelligent prediction method can discover and solve potential problems in advance, optimize resource allocation, and improve the stability and efficiency of the system.

[0034] 2. Improvement of actual application value

[0035] The output of predictive analysis results is not limited to simple data reports, but also includes detailed information such as problem descriptions and solution suggestions; this information can be directly used to guide the actual work of operation and maintenance personnel, helping them to locate and solve faults more quickly and schedule resources more reasonably; by storing these prediction results in the database and performing subsequent queries and statistical applications, the value of the data can be further mined, providing a basis for continuous optimization and improvement of the system.

[0036] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0038] Figure 1 This is a flow chart of the method of the first embodiment.

[0039] Figure 2 This is a structural diagram of the overall fault prediction model of the first embodiment.

[0040] Figure 3 This is a structural diagram of a specific fault type prediction model of the first embodiment.

[0041] Figure 4 This is a structural diagram of the anomaly detection model of the first embodiment. DETAILED DESCRIPTION

[0042] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] Embodiment 1

[0045] In one embodiment of the present disclosure, an operation and maintenance assistance method applicable to an information system of a human resources and social security department is provided. Figure 1 As shown, the following steps are included:

[0046] Step S1: Acquire the real-time collected operation and maintenance data of the human resources and social security department information system and perform preprocessing;

[0047] Furthermore, the preprocessing includes removing duplicate, erroneous, and irrelevant data, removing noise, standardizing, and extracting features.

[0048] Furthermore, the feature extraction is to extract the features that characterize the status of the operating system, database, middleware and application system respectively, specifically:

[0049] Use file system usage, system and network log alarm information, CPU load, storage I / O performance, date, and time to represent the operating system status;

[0050] The database status is represented by database health status, ASM disk group and disk information, data file storage space information, table space usage, disaster recovery operation information, backup and archive log storage information, and monitoring logs.

[0051] Use node instance running status, log storage space information, and thread blocking information to represent the middleware status;

[0052] The application system status is represented by access information, system operation status information, database user memory usage, process waiting information, database lock information, microservice logs, API call data, and user behavior logs.

[0053] Step S2: Input the preprocessed operation and maintenance data into the trained model to predict future fault occurrence and resource requirements, so as to assist in system fault handling and resource scheduling;

[0054] Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

[0055] Furthermore, the model includes a fault prediction model and a resource prediction model;

[0056] The fault prediction model uses the time series of operation and maintenance data as input to predict the probability of faults and the type of faults;

[0057] The resource prediction model uses operation and maintenance data as input and is used to predict future file system usage, future CPU usage, and future memory usage.

[0058] Furthermore, the fault prediction model adds a softmax layer to perform classification on the basis of RNN and outputs the predicted probability of each fault type.

[0059] Furthermore, the resource prediction model uses a gradient boosting tree algorithm to construct a regression model to predict the file system usage, CPU usage, and memory usage at future time points.

[0060] Furthermore, the model also includes using an unsupervised learning model to detect abnormal data.

[0061] This embodiment provides a prediction method based on artificial intelligence and operation and maintenance data, which realizes the prediction application of faults and resources in the case of deploying the human resources and social security LEAF framework application in the government cloud with IAAS capabilities; this new prediction method based on artificial intelligence and operation and maintenance data needs to be widely applicable to the operation and maintenance of the information system of the human resources and social security departments, and have safe, stable and widely applicable technical characteristics to improve the speed of fault location and efficient resource management, effectively reduce operation and maintenance costs, and build a solid technical support for improving service efficiency and user experience.

[0062] The following describes in detail the implementation process of an operation and maintenance assistance method suitable for an information system of a human resources and social security department in this embodiment from the aspects of operation and maintenance data collection, operation and maintenance data storage and management, operation and maintenance data processing, fault prediction model management, resource prediction model management, and application of prediction and analysis results.

[0063] 1. Operation and maintenance data collection

[0064] Comprehensively collect multi-dimensional data from IaaS government cloud infrastructure (such as virtual machine performance, network status, storage I / O, etc.), LEAF6 framework applications of human resources and social security (such as microservice logs, API call data, user behavior logs, etc.) and third-party monitoring systems (such as security monitoring, performance monitoring, etc.); use timed polling, event triggering, stream processing and other methods to ensure the real-time and integrity of data, specifically:

[0065] (1) Define data collection strategy

[0066] Identify data sources: First, identify and list all data sources that need to be monitored, including but not limited to operating systems, databases, middleware, application systems, etc.

[0067] Data type and frequency: Determine the type of data to be collected (such as performance indicators, logs, alarms, etc.) and the frequency of collection (such as every second, every minute, every hour, etc.) based on the importance and real-time requirements of each data source.

[0068] (2) Data collection through timed polling

[0069] Use Shell scripts and Python scripts combined with top, vmstat, df, free, iostat and other system commands to collect operating system level data.

[0070] Use SQL scripts to regularly query the database to obtain database health status, disk group information, table space usage, etc.

[0071] Use the API or management interface provided by the middleware to regularly query the node instance status, log storage space, etc.

[0072] (3) Data collection through event triggering

[0073] Use log collection tools such as Logstash and Fluentd to monitor system and application log files to collect log data in real time or quasi real time.

[0074] Integrate the existing alarm system to automatically trigger data collection when the system generates an alarm.

[0075] (4) Data collection through stream processing

[0076] Configure stream processing tools such as Kafka or Flume at the log generation or middleware output end to capture and forward data streams to the data center or analysis system in real time.

[0077] 2. Operation and maintenance data storage and management

[0078] Use distributed databases or big data platforms to store massive operation and maintenance data; establish efficient data indexing and query mechanisms to speed up data retrieval; use encrypted storage, access control and other measures to ensure data security, specifically:

[0079] (1) Data aggregation and storage

[0080] Aggregate the collected data into a database; for scenarios that require high-performance reading, writing, and storage of time series data (such as performance indicators), use a time series database (TSDB) for storage.

[0081] (2) Data quality monitoring

[0082] Use statistical methods and machine learning algorithms to detect abnormal data and promptly identify and address problems in data collection.

[0083] (3) Data encryption

[0084] Encryption technology is used during data transmission and storage to ensure data security.

[0085] 3. Operation and Maintenance Data Processing

[0086] Remove duplicate, erroneous, and irrelevant data, process missing values, and reduce noise interference; standardize the data to eliminate the impact of different dimensions and improve data comparability; extract features that have an important impact on fault prediction and resource prediction from the preprocessed data, specifically:

[0087] (1) Remove duplicate, erroneous, and irrelevant data

[0088] Remove duplicate data: Use database query statements (such as the DISTINCT keyword or GROUP BY statement in SQL) or programming languages ​​(such as the drop_duplicates() method of the pandas library in Python) to identify and remove duplicate records.

[0089] Identify and handle incorrect data: Incorrect data includes format errors, logical errors (such as negative age), or range errors (such as dates within an impossible time range). Error criteria need to be manually defined based on business rules, and then identified and corrected using scripts or data cleaning tools.

[0090] Remove irrelevant data: Based on business needs and data analysis goals, identify which fields or records are not required for analysis and then remove them from the dataset.

[0091] (2) Reduce noise interference

[0092] Smoothing: For time series data, use methods such as moving average and exponential smoothing to reduce random noise.

[0093] Outlier detection and processing: Use statistical methods (Z-score, IQR, etc.) or machine learning algorithms (isolation forest) to detect outliers and choose to retain, correct, or delete them based on business needs.

[0094] (3) Data standardization

[0095] Convert the data to a distribution with a mean of 0 and a standard deviation of 1; eliminate the influence of different dimensions and make the data distribution more even for subsequent analysis.

[0096] (4) Feature extraction, see Table 1 for details

[0097] Table 1 Extracted feature table

[0098]

[0099]

[0100] 4. Fault prediction model management

[0101] Build a fault prediction model of the LEAF framework based on machine learning algorithms; use historical fault data and feature data to train the model and optimize model parameters; evaluate model performance through cross-validation, A / B testing and other methods to ensure prediction accuracy, specifically:

[0102] (1) Constructing a fault prediction model based on the LEAF framework

[0103] Taking into account factors such as the type of data (numerical, categorical, time series, etc.), the dimension of the data (number of features), the distribution of the data (such as whether it is skewed), the nature of the target task (classification, regression, clustering, time series prediction, etc.), and the size of the data set, the following models are included for three different objectives:

[0104] 1) Overall failure prediction

[0105] Predict whether the system will fail (binary classification problem), use RNN variants such as LSTM or GRU combined with a fully connected layer (Dense Layer) for classification, and capture the long-term dependencies in the time series through the model to improve the accuracy of the prediction. The model structure is as follows Figure 2 As shown, including:

[0106] Input layer (Input): Receives the time series of feature vectors as input. The data set is constructed by the number of features × time step (per minute). The number of features refers to the features in Table 1, and the specific configuration is based on the actual situation. The model is input according to the aggregated data per minute (the corresponding data of each feature is generally in milliseconds).

[0107] Data processing layer (LSTM / GRU Layer 1 and LSTM / GRU Layer 2): Use RNN variants such as LSTM or GRU to capture long-term dependencies in time series data; stack 2 LSTM / GRU layers to increase model depth and improve learning ability. The parameter of the number of Units neurons in each layer is adjusted appropriately based on a comprehensive evaluation of the dataset size and computing resources. In the overall fault prediction model, 2 layers of LSTM / GRU are sufficient, and further increasing the number of layers will only result in limited performance improvement.

[0108] Feature Integration Layer (DenseLayer): After the data processing layer, a feature integration layer (Dense Layer) is added to further process feature data. For example, the CPU load of the operating system, memory usage, table space usage of the database, thread blocking information of the middleware, etc. are located at different levels and are integrated together in the fully connected layer to form a comprehensive understanding of the overall health of the system. This layer uses the activation function ReLU to analyze the data distribution pattern of the feature data of each component of the learning system (such as CPU load, memory usage, database health status, etc.) under normal conditions. When the actual data deviates significantly from this pattern, the model can identify potential abnormal behavior.

[0109] Output Layer: Use the Sigmoid activation function to compress the output of the neuron (i.e., the weighted sum plus the bias term) into the (0,1) interval to obtain a probability value, which indicates the possibility of system failure. If the output value is close to 0, it indicates that the possibility of system failure is very low; if the output value is close to 1, it indicates that the possibility of system failure is very high. After setting the system failure threshold (such as 0.7) and comparing it with the predicted probability value generated by the Sigmoid activation function, the output indicates whether the system has failed.

[0110] Loss function: During the training process, the binary cross entropy loss function (Binary CrossentropyLoss) is used to evaluate the difference between the model's predicted value and the actual value, calculate the difference between the probability distribution predicted by the model and the actual label (0 or 1), and return a scalar value to represent the size of this difference. During the training process, the optimizer will try to minimize this loss value, thereby adjusting the model's weights to make the model's predictions more accurate.

[0111] 2) Prediction of specific fault types

[0112] To predict the specific fault type (multi-classification problem), add a softmax layer to the RNN for classification; at the same time, use the attention mechanism to enhance the model's attention to key information. The model structure is as follows: Figure 3 As shown, specifically including:

[0113] Input layer (Input): Receives the time series of feature vectors as input. The data set is constructed by the number of features × time step (per minute). The number of features refers to the features in Table 1, and the specific configuration is based on the actual situation. The model is input according to the aggregated data per minute (the corresponding data of each feature is generally in milliseconds).

[0114] Data processing layer (LSTM / GRU Layer 1 and LSTM / GRU Layer 2): Use RNN variants such as LSTM or GRU to capture long-term dependencies in time series data; stack 2 LSTM / GRU layers to increase model depth and improve learning ability. The parameter of the number of Units neurons in each layer is adjusted appropriately based on a comprehensive evaluation of the dataset size and computing resources. In the overall fault prediction model, 2 layers of LSTM / GRU are sufficient, and further increasing the number of layers will only result in limited performance improvement.

[0115] Attention Layer: Dynamically calculate the weight of each time step based on the output of the data processing layer, and normalize these scores into probability distribution through the softmax function as the final attention weight.

[0116] Feature Integration Layer (DenseLayer): After the attention layer, add a feature integration layer (Dense Layer) to further process the feature data. Further integrate and process the attention-weighted features. Through the activation function ReLU, analyze the data distribution pattern of the feature data of each component of the learning system (such as CPU load, memory usage, database health status, etc.) under normal conditions. When the actual data deviates significantly from this pattern, the model can identify potential abnormal behavior.

[0117] Softmax layer (Output Layer): performs multi-classification output. It receives the output of the feature integration layer and converts it into a set of probability values, each of which corresponds to a possible fault type.

[0118] Loss function: Use categorical cross entropy loss as the loss function to evaluate the difference between the probability distribution predicted by the model and the true label. During the training process, the optimizer will try to minimize this loss value, thereby adjusting the weights of the model to make the model's predictions more accurate.

[0119] 3) Anomaly Detection

[0120] There are a large number of normal samples in the data but few abnormal samples, so an autoencoder is used for anomaly detection; the model learns the distribution of normal data and identifies abnormal samples that deviate greatly from the normal pattern. The model structure is as follows: Figure 4 As shown, specifically including:

[0121] Input layer: Receives characteristic data samples (such as operating systems, databases, middleware, application systems, etc.). Data samples are organized into feature vectors, each of which contains multiple features, such as CPU usage, memory usage, database table space usage, etc.

[0122] Encoder (Encoder Layer 1…Encoder Layer n): Use the ReLU activation function to process the input data and map it to a low-dimensional hidden representation. By stacking hidden layers that use the ReLU activation function, the encoder gradually extracts high-level features of the input data and removes redundant information to achieve dimensionality reduction and compression of the data. The ReLU function maps the input value to the interval [0,+∞]. If the input is less than 0, the output is 0; otherwise the output is equal to the input. Each layer of the encoder receives the output of the previous layer as input, performs a linear transformation through the weight matrix and bias term, and then applies the ReLU activation function for nonlinear transformation. This process is repeated until the bottleneck layer is reached.

[0123] Bottleneck Layer: Receives the output of the last layer of the encoder and compresses it into a low-dimensional vector. This vector is an efficient representation of the input data and contains the key information needed to reconstruct the input data.

[0124] Decoder: Maps the low-dimensional representation of the bottleneck layer back to the high-dimensional space of the original data to reconstruct the input data. Use the Tanh activation function to process continuous data (such as CPU usage) and map the output to the [-1,1] interval, which helps to keep the data centered. Use the Sigmoid activation function to process categorical data and map the output to the [0,1] interval, which can be interpreted as a probability.

[0125] Output Layer: Receives the output of the last layer of the decoder and produces a reconstructed data sample. This sample is then compared with the original input data to calculate the reconstruction error (such as mean square error MSE), which is used to train the model and identify abnormal samples. During the training process, the model learns the distribution of normal data and is able to identify abnormal samples that deviate greatly from the normal pattern through the reconstruction error.

[0126] (2) Fault prediction model training and parameter optimization

[0127] Use optimizers (such as Adam, RMSprop) for training, monitor the loss and validation metrics (such as accuracy, recall, F1 score) during training to avoid overfitting; reduce overfitting through techniques such as early stopping, dropout, regularization, etc.

[0128] (3) Model evaluation

[0129] Evaluate model performance on an independent test set; adjust model parameters and architecture based on business needs; deploy the model to a production environment and continuously monitor its performance, retraining or adjusting when necessary.

[0130] 5. Resource Forecast Model Management

[0131] Build a resource prediction model for the LEAF framework based on machine learning algorithms; use storage, memory, and CPU growth trend feature data to train the model and optimize model parameters; evaluate model performance through cross-validation, A / B testing, and other methods to ensure prediction accuracy, specifically:

[0132] (1) Constructing a resource forecasting model based on the LEAF framework

[0133] The gradient boosting tree (such as XGBoost, LightGBM) algorithm is used to build a regression model to predict the file system usage, CPU usage, memory usage, etc. at a certain point in the future.

[0134] Input layer: The input layer contains all the features used for prediction.

[0135] Processing layer:

[0136] Feature conversion: Convert features such as date and time into a format suitable for model processing (such as decomposing timestamps into years, months, days, hours, etc.).

[0137] Feature selection: Select the most useful features through feature importance analysis or dimensionality reduction techniques.

[0138] Tree splitting: The model builds multiple decision trees by continuously selecting the best features and splitting points; each tree tries to reduce the prediction error.

[0139] Ensemble learning: The prediction results of all trees are weighted averaged to obtain the final prediction value.

[0140] Output layer: outputs prediction results, including: future file system usage prediction, future CPU usage prediction, future memory usage prediction, etc.

[0141] (2) Resource prediction model training and parameter optimization

[0142] Use the processed data to train the model; during the training process, it is necessary to set appropriate model parameters (such as learning rate, batch size, number of iterations, etc.) and adjust the model parameters through optimization algorithms such as gradient descent.

[0143] Optimize model performance through hyperparameter tuning; hyperparameter tuning can adopt grid search, random search, Bayesian optimization and other methods; these methods can automatically explore model performance under different parameter combinations and find the optimal parameter configuration.

[0144] (3) Model evaluation

[0145] Use cross-validation to evaluate the generalization ability of the model; divide the data set into multiple subsets, take each subset as the test set in turn, and use the remaining subsets as the training set for training and testing; through multiple training and testing, the performance of the model on different data can be evaluated more accurately.

[0146] Conduct A / B testing to evaluate the performance of the model; compare the results predicted by the model with the results of existing systems or other models, and evaluate the accuracy of the model by comparing the difference between actual resource growth and predicted results.

[0147] 6. Application of Predictive Analysis Results

[0148] The predictive analysis results include information such as problem description and solution suggestions. After generation, they are stored in the database for subsequent query and statistical application. Operation and maintenance personnel can use the results to assist in troubleshooting and resource scheduling.

[0149] 1. Fault prediction model analysis results

[0150] (1) Failure Cause Analysis

[0151] Historical fault data correlation: Analyze the correlation between historical fault data and current system status (such as operating system status, file system usage, CPU load, memory usage, etc.) to identify which characteristics are most likely to cause specific types of faults.

[0152] Alarm information analysis: Analyze system and network log alarm information, especially error and warning level alarms, and infer possible fault sources based on alarm type, occurrence time, and system status changes.

[0153] Time series analysis: Use time series data (such as CPU load, memory usage, access volume changes, etc.) to analyze the trend and periodicity of failures and predict possible future failure time windows.

[0154] (2) Application of prediction results

[0155] Early warning: Based on the prediction results, an early warning is issued to operation and maintenance personnel before the system reaches the failure threshold, providing the time when the failure may occur, the scope of impact and the possible causes.

[0156] Contingency plan: Prepare contingency plans in advance based on the prediction results, including backing up data, adjusting resource allocation, fixing potential software defects, etc.

[0157] 2. Resource prediction model analysis results

[0158] (1) Resource Demand Forecast

[0159] Trend analysis: Use storage, memory, and CPU growth trend feature data to predict future resource requirements.

[0160] Business load correlation: Correlate and analyze resource usage with business load (such as access volume, database user memory usage, etc.) to understand the impact of business needs on resource needs.

[0161] Resource bottleneck identification: Analyze current resource allocation and usage, and identify potential resource bottlenecks, such as insufficient storage space, CPU or memory overload, etc.

[0162] (2) Resource allocation optimization

[0163] Dynamic resource scheduling: Based on the prediction results, dynamic resource scheduling is implemented, such as automatic expansion of storage capacity, increase of CPU or memory resources, etc., to cope with future business growth.

[0164] Cost optimization: Under the premise of ensuring stable operation of the system, reduce resource waste and lower operating costs through reasonable resource allocation.

[0165] Embodiment 2

[0166] In one embodiment of the present disclosure, an operation and maintenance assistance system applicable to an information system of a human resources and social security department is provided, including an acquisition module and a prediction module:

[0167] The acquisition module is configured to: acquire the real-time collected operation and maintenance data of the information system of the human resources and social security department and perform pre-processing;

[0168] The prediction module is configured to: input the pre-processed operation and maintenance data into the trained model to predict future fault occurrence and resource requirements to assist in system fault handling and resource scheduling;

[0169] Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

[0170] Embodiment 3

[0171] The purpose of this embodiment is to provide a computer-readable storage medium.

[0172] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of an operation and maintenance assistance method applicable to an information system of a human resources and social security department as described in Embodiment 1 of the present disclosure.

[0173] Embodiment 4

[0174] The purpose of this embodiment is to provide an electronic device.

[0175] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of an operation and maintenance assistance method applicable to an information system of a human resources and social security department as described in Embodiment 1 of the present disclosure are implemented.

[0176] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An operation and maintenance auxiliary method suitable for the information system of human resources and social security departments, characterized in that: include: Obtain real-time collected operation and maintenance data of the human resources and social security department's information system and perform pre-processing; The pre-processed operation and maintenance data is input into the trained model to predict future fault occurrence and resource requirements, which is used to assist in system fault handling and resource scheduling. Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

2. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 1, characterized in that: The preprocessing includes removing duplicate, erroneous, and irrelevant data, removing noise, standardizing, and extracting features.

3. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 2, characterized in that: The feature extraction is to extract the features that characterize the status of the operating system, database, middleware and application system respectively, specifically: Use file system usage, system and network log alarm information, CPU load, storage I / O performance, date, and time to represent the operating system status; The database status is represented by database health status, ASM disk group and disk information, data file storage space information, table space usage, disaster recovery operation information, backup and archive log storage information, and monitoring logs. Use node instance running status, log storage space information, and thread blocking information to represent the middleware status; The application system status is represented by access information, system operation status information, database user memory usage, process waiting information, database lock information, microservice logs, API call data, and user behavior logs.

4. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 1, characterized in that: The model includes a fault prediction model and a resource prediction model; The fault prediction model uses the time series of operation and maintenance data as input to predict the probability of faults and the type of faults; The resource prediction model uses operation and maintenance data as input and is used to predict future file system usage, future CPU usage, and future memory usage.

5. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 4, characterized in that: The fault prediction model adds a softmax layer to the RNN for classification and outputs the predicted probability of each fault type.

6. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 4, characterized in that: The resource prediction model uses a gradient boosting tree algorithm to build a regression model to predict the file system usage, CPU usage, and memory usage at a future time point.

7. The operation and maintenance auxiliary method applicable to the information system of the human resources and social security department as claimed in claim 1, characterized in that: The model also includes using an unsupervised learning model to detect abnormal data.

8. An operation and maintenance auxiliary system suitable for the information system of the human resources and social security department, characterized in that: Including acquisition module and prediction module: The acquisition module is configured to: acquire the real-time collected operation and maintenance data of the information system of the human resources and social security department and perform pre-processing; The prediction module is configured to: input the pre-processed operation and maintenance data into the trained model to predict future fault occurrence and resource requirements to assist in system fault handling and resource scheduling; Among them, the operation and maintenance data is collected in real time by combining timed polling, event triggering and stream processing, including virtual machine performance data, network status data, storage data, microservice log data, API call data, user behavior log data, security monitoring data, and performance monitoring data.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; a processor for executing the computer-readable instructions; Wherein, when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitorily stored, wherein when the computer-readable instructions are executed by a computer, the method of any one of claims 1 to 7 is performed.

Images