Information and credential equipment maintenance method, system and equipment and storage medium

Abstract

The invention provides a credential equipment maintenance method, system and equipment and a storage medium, and relates to the technical field of operation and maintenance. According to the method, for the problem of low multi-source heterogeneous data processing efficiency, semantic importance evaluation is carried out on collected data through a service weight distribution algorithm, and the data sorting priority is dynamically adjusted in combination with a BM25 weight value fusion technology, so that key service information is highlighted in operation and maintenance monitoring, and the continuity of high-priority services is ensured; aiming at the problem of abnormal data recognition lag in a maintenance scene, a comprehensive weight index model is constructed through personnel behavior feature analysis (access times, diversity and repeatability), the abnormal behavior risk of personnel entering and exiting from a spare part library is quantified, and the real-time recognition capability of abnormal data is improved. According to the invention, intelligentization, precision and high efficiency of operation and maintenance integration can be realized, and the business continuity and equipment reliability of the data center are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance technology, and in particular to a method, system, equipment and storage medium for the maintenance of information technology application innovation equipment. Background Technology

[0002] In the operation and maintenance scenarios of intelligent agent applications in the field of information technology innovation, existing technologies typically employ multi-source heterogeneous data collection and classification, equipment fault prediction based on empirical rules, and data analysis methods primarily based on manual monitoring. Traditional operation and maintenance systems prioritize collected data through fixed weight allocation or simple threshold judgment, making it difficult to dynamically respond to real-time changes in key business semantic information, resulting in insufficient continuity assurance for high-priority businesses in operation and maintenance monitoring.

[0003] In the maintenance process, existing technologies mostly rely on manual inspections or static rule bases to detect anomalies in spare parts warehouse data. They lack quantitative analysis of personnel behavior characteristics and make it difficult to identify potential risk behaviors such as unauthorized operations and abnormal access in a timely manner.

[0004] For fault prediction of domestically developed IT equipment, existing methods are usually based on a single empirical model (such as historical failure rate statistics) to make judgments. They cannot combine the dynamic changes in the operating status of the equipment and the correlation of multi-dimensional parameters in complex scenarios, resulting in insufficient prediction accuracy, high equipment startup failure rate (usually exceeding 10%), and lengthy debugging cycle, which directly affects the stability and economy of business systems.

[0005] Therefore, how to solve the problems of low efficiency in processing multi-source heterogeneous data, ambiguous priority of key business semantic information, insufficient accuracy in equipment fault prediction, and lag in abnormal behavior identification in the integrated operation and maintenance of intelligent agent applications in the field of information technology innovation has become an urgent technical challenge. Summary of the Invention

[0006] Purpose of the invention: To propose a method, system, equipment, and storage medium for the maintenance of information technology application innovation (ITAI) equipment, in order to solve the problems mentioned in the background art.

[0007] In a first aspect, the present invention provides a maintenance method for information technology application innovation (ITAI) equipment, comprising the following steps: S1. Collect and classify multi-source heterogeneous data from the production environment consisting of customer data center IT systems, networks, servers, and software; use a business weight allocation algorithm to assign weights to the collected data of each embedded model based on the semantic importance of the collected data, and then merge the weight values ​​with the traditional BM25 weight values ​​to obtain a dataset. S2. Mark the devices corresponding to the collected data according to the business weights obtained in step S1, and store the marked devices in the spare parts warehouse; analyze the behavior of personnel entering and leaving the spare parts warehouse, extract three types of features: number of visits, access diversity, and repeated visits, calculate the behavior weight of each person entering and leaving the warehouse in each application scenario, and then obtain the comprehensive weight index of each person entering and leaving the warehouse, and judge whether the data entering and leaving the spare parts warehouse is abnormal based on the comprehensive weight index. S3. For the devices corresponding to the abnormal data in and out of the spare parts warehouse identified in step S2, use the Monte Carlo method to simulate the probability of device failure, combine computing power and efficiency indicators to evaluate device performance, identify abnormal device startup behavior, and ensure that the domestically developed equipment is in normal working condition.

[0008] In a further embodiment of the first aspect, in step S1, the semantic importance of the collected data is evaluated using natural language processing techniques, including word frequency-inverse document frequency (TF-IDF) and word vector similarity. The formula for calculating the weight assignment is:

[0009] in, These are the weights of the collected data for the i-th embedded model. is the semantic importance score of the collected data of the i-th embedding model, and f is the mapping function that converts the importance score into a weight value.

[0010] In a further embodiment of the first aspect, the calculation formula for weight value fusion in step S1 is:

[0011] Where d represents the collected classification dataset, and n is the number of weights in the embedding model. and These are the weights of the i-th embedded model and the weights of BM25, respectively. and These are the weight scores of the i-th embedding model and BM25, respectively.

[0012] In a further embodiment of the first aspect, the formula for calculating the comprehensive weight index, i.e., the data center access content index, in step S2 is as follows:

[0013] Where M represents the content accessed by the data center, U represents the set of people entering and leaving the data center, and W(u) represents the behavioral weight of the people entering and leaving the data center. The smaller the value, the greater the probability that the person entering or leaving the data center is abnormal; the larger the value, the greater the probability that the person entering or leaving the data center is a high-value user. This is the formula for calculating the content access index for people entering and leaving traditional data centers.

[0014] In a further embodiment of the first aspect, the formula for simulating the device failure probability using the Monte Carlo method in step S3 is:

[0015] in, It is the failure probability of the i-th component. This indicates a product, where at least one component in the system is guaranteed to fail, meaning at least one... At that time, the failure probability of the entire system .

[0016] In a further embodiment of the first aspect, the calculation formula for computing power CP in step 3 is:

[0017] in, It refers to general-purpose computing power, represented by the central processing unit (CPU). High-performance computing power, represented by graphics processing units (GPUs), is commonly measured in EFLOPS. The formula for calculating computational efficiency (CE) is: Where CP represents the computing power of the data center, expressed as a single-precision floating-point number FP32, and PCT represents the overall power of the data center IT equipment, in watts (W).

[0018] A second aspect of the present invention provides a maintenance system for information technology application innovation (ITAI) equipment, the system comprising: The operation and maintenance module is used to collect and classify multi-source heterogeneous data in the production environment consisting of customer data center IT systems, networks, servers, and software. It adopts a business weight allocation algorithm to assign weights to the collected data of each embedded model according to the semantic importance of the collected data, and then merges the weight values ​​with the traditional BM25 weight values ​​to obtain the data set. The maintenance module is used to mark the equipment corresponding to the collected data according to the business weight obtained by the operation and maintenance module, and store the marked equipment in the spare parts warehouse; analyze the behavior of personnel entering and leaving the spare parts warehouse, extract three types of features: number of visits, access diversity, and repeated access, calculate the behavior weight of each person entering and leaving the warehouse in each application scenario, and then obtain the comprehensive weight index of each person entering and leaving the warehouse, and judge whether the data entering and leaving the spare parts warehouse is abnormal based on the comprehensive weight index. The IT innovation module is used to simulate the probability of equipment failure using the Monte Carlo method for the equipment corresponding to the abnormal data of the spare parts warehouse identified by the maintenance module, and to evaluate the equipment performance by combining computing power and efficiency indicators, and to identify abnormal equipment startup behavior.

[0019] In a third aspect, the present invention provides an electronic device comprising: a processor and a memory storing computer program instructions; wherein the processor, when executing the computer program instructions, implements the maintenance method for domestically developed information technology equipment disclosed in the first aspect and its further embodiments.

[0020] In a fourth aspect, the present invention provides a computer-readable storage medium storing at least one executable instruction that, when executed on an electronic device, causes the electronic device to perform the maintenance method for domestically developed innovative equipment disclosed in the first aspect and its further embodiments.

[0021] Beneficial effects: (1) This invention dynamically evaluates the semantic importance of data by combining a business weight allocation algorithm with natural language processing technology and integrates it with BM25 weight values ​​to achieve intelligent priority sorting of multi-source heterogeneous data, ensuring the real-time prominence of key business information. This solves the problems of low processing efficiency of multi-source heterogeneous data and ambiguous priority of key business semantic information in the prior art, and ensures the continuity of high-priority business.

[0022] (2) The present invention constructs a personnel behavior characteristic analysis model, which quantifies the abnormal behavior risk of personnel entering and leaving the spare parts warehouse through three types of indicators: number of visits, diversity and repetition, improves the real-time and accuracy of abnormal data identification, solves the problem of lagging abnormal behavior identification in maintenance scenarios, strengthens spare parts warehouse management, and ensures the security of the information technology innovation equipment supply chain.

[0023] (3) The present invention adopts a method that combines Monte Carlo fault simulation with computing power and efficiency index evaluation to dynamically simulate the probability of equipment failure and optimize performance parameters, which solves the problem of insufficient accuracy in predicting equipment failure in the field of information technology innovation and significantly improves business continuity and equipment reliability in the field of information technology innovation.

[0024] (4) The present invention divides the system into an operation and maintenance module, a maintenance module and an information technology innovation module. The high-weight data output by the operation and maintenance module directly drives the equipment marking and spare parts library management of the maintenance module. The personnel behavior analysis results of the maintenance module are fed back to the operation and maintenance module to form a closed-loop optimization. Data linkage between modules reduces redundant processing and improves the overall operation and maintenance efficiency. Dynamic parameter tuning enables the system to adapt to the needs of data centers of different sizes (such as small computer rooms and ultra-large-scale data centers). Attached Figure Description

[0025] Figure 1 This is an architecture diagram of the maintenance system for domestically developed equipment in this embodiment. Detailed Implementation

[0026] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0027] This embodiment discloses a maintenance method for a reliable and innovative device, including the following steps: Step 1: The operation and maintenance module collects and classifies multi-source heterogeneous data from the production environment, which consists of customer data center IT systems, networks, servers, and software. It adopts a business weight allocation algorithm to assign weights to the collected data of each embedded model based on the semantic importance of the collected data. Then, it merges the weight values ​​with the traditional BM25 weight values ​​to obtain a data set. Based on this data set, it realizes continuous operation monitoring and management of the production environment to ensure uninterrupted business.

[0028] The semantic importance of the collected data is evaluated using natural language processing (NLP) techniques, including Term Frequency-Inverse Document Frequency (TF-IDF) and word vector similarity. For example, when using TF-IDF, for the collected text data, the frequency of each word in the text (TF) and the document frequency of the word in the entire dataset (IDF) are calculated. The importance index of the word is obtained by multiplying TF and IDF, thereby comprehensively evaluating the semantic importance of the entire text data. When using word vector similarity, the words in the collected data are converted into vector form, and the similarity between different word vectors is calculated. The degree of semantic association between words is determined based on the similarity results, thereby evaluating the semantic importance of the data.

[0029] The formula for calculating the weight assignment is: ,in, These are the weights of the collected data for the i-th embedded model. Let f be the semantic importance score of the collected data for the i-th embedded model, and let f be the mapping function that converts the importance score into a weight value. The mapping function f can be set according to the actual application scenario. For example, when the importance score is 0-20, the mapping function f can convert it into a weight value of 0.1-0.3; when the importance score is 21-50, it can be converted into a weight value of 0.4-0.6; and when the importance score is 51-100, it can be converted into a weight value of 0.7-1.0, so as to achieve reasonable weight allocation according to semantic importance.

[0030] The formula for calculating the weighted value fusion is: Where d represents the collected classification dataset, and n is the number of weights in the embedding model. and These are the weights of the i-th embedded model and the weights of BM25, respectively. and These are the weight scores of the i-th embedding model and BM25, respectively. In practical applications, the weights can be set according to the importance of different embedding models and BM25 in data processing. and The value, for example, if a certain embedding model performs better when processing a specific type of data (such as server log data), can be set to its corresponding value. The value is set to 0.3, while BM25's The value is set to 0.4, for other embedded models. The values ​​are set separately based on their performance, and all and The sum of the values ​​is 1 to ensure the rationality of weight fusion.

[0031] Step 2: The maintenance module marks the devices corresponding to the collected data according to the business weights obtained in Step 1, and stores the marked devices in the spare parts warehouse; it analyzes the behavior of personnel entering and leaving the spare parts warehouse, extracts three types of features: number of visits, access diversity, and repeat visits, calculates the behavioral weight of each person entering and leaving the warehouse in each application scenario, and then obtains the comprehensive weight index of each person entering and leaving the warehouse. Based on the comprehensive weight index, it determines whether the data entering and leaving the spare parts warehouse is abnormal.

[0032] The access frequency characteristic refers to the total number of times personnel enter and exit the spare parts warehouse within a set time period (such as a day or a week). For example, if personnel A enters and exits the spare parts warehouse 10 times and personnel B enters and exits 20 times in a week, the frequency of personnel activities can be preliminarily understood by counting this number.

[0033] Access diversity refers to the degree of diversity of warehouse areas and equipment types involved when personnel enter and exit the spare parts warehouse. For example, personnel A only moves in the server storage area of ​​the spare parts warehouse, and the equipment type involved is only servers; while personnel B moves in both the server storage area and the network equipment storage area, involving multiple equipment types such as servers, switches, and routers. Personnel B's access diversity is higher than that of personnel A.

[0034] Repeatability refers to the number of times a person repeatedly accesses the same warehouse area or the same equipment within a short period of time (e.g., within 1 hour). For example, if person C repeatedly accesses a specific server location in the spare parts warehouse 3 times within 1 hour, while person D does not repeatedly access the same location within the same time period, then person C's repeatability is higher than that of person D.

[0035] The formula for calculating the comprehensive weighted index, also known as the data center access content index, is as follows: Where M represents the content accessed by the data center, U represents the set of people entering and leaving the data center, and W(u) represents the behavioral weight of the people entering and leaving the data center. The smaller the value, the greater the probability that the person entering or leaving the data center is abnormal; the larger the value, the greater the probability that the person entering or leaving the data center is a high-value user. This is a traditional formula for calculating the content access index for data center users, related to data such as clicks, comments, reposts, and likes. For example, for user E, their behavior weight W(u) is calculated to be 0.8, which is the traditional content access index. If the value is 0.6, then the person's comprehensive weight index is... If the behavioral weight W(u) of person F is 0.2, the traditional access content index If the value is 0.5, then its comprehensive weight index is... Personnel F has a low overall weight index, which suggests a high possibility of anomalies. Further monitoring of their entry and exit from the spare parts warehouse is required.

[0036] Step 3: For the network devices, servers, hardware and other IT innovation devices corresponding to the abnormal data in and out of the spare parts warehouse identified in Step 2, the IT innovation module uses the Monte Carlo method to simulate the probability of device failure, combines computing power and efficiency indicators to evaluate device performance, identify abnormal device startup behavior, and ensure that the IT innovation devices are in normal working order.

[0037] The formula for simulating equipment failure probability using the Monte Carlo method is: ,in, It is the failure probability of the i-th component. This indicates a product, where at least one component in the system is guaranteed to fail, meaning at least one... At that time, the failure probability of the entire system For example, a certain domestically developed IT equipment consists of three main components: CPU, memory, and hard drive, with failure probabilities of […]. , , The failure probability of the device is... That is, a failure probability of approximately 5.9%.

[0038] The formula for calculating computing power (CP) is: ,in, It refers to general-purpose computing power, represented by the central processing unit (CPU). High-performance computing power, represented by graphics processing units (GPUs), is commonly measured in EFLOPS (Image Processing Units). (FLOPS). For example, the general computing power provided by the CPUs in a data center. EFLOPS, the high-performance computing power provided by GPUs EFLOPS represents the computing power of the data center. EFLOPS.

[0039] The formula for calculating the computational efficiency (CE) is: Where CP represents the computing power of the data center, expressed as a single-precision floating-point number FP32, and PCT represents the total power of the data center's IT equipment, in watts (W). For example, if a data center has a computing power CP = 0.8 EFLOPS and a total power PCT = 10000W, then the computing efficiency of this data center is... FLOPS / 10000W = FLOPS / W. By using computing power and computing efficiency indicators, the computing power and energy efficiency of domestically developed IT equipment can be evaluated. If the computing power or computing efficiency is too low, it may indicate that there are performance problems or potential faults in the equipment, which require further investigation.

[0040] This embodiment provides a maintenance method for trusted IT innovation equipment, applied to a large enterprise's IT innovation data center. This data center contains multiple computer rooms and deploys a large number of servers, network devices, and other IT facilities. The specific implementation steps are as follows: I. Data Acquisition and Processing in the Operation and Maintenance Module Data Acquisition: The operations and maintenance module uses sensors and data acquisition agents deployed in various data centers to collect multi-source heterogeneous data from the production environment, consisting of customer data center IT systems, networks, servers, and software. The collected data types include server operation logs (such as CPU utilization, memory usage, disk I / O, etc.), network traffic data (such as bandwidth utilization, packet loss rate, etc.), and software application logs (such as error logs, access logs, etc.). The collection frequency is set to once every 5 minutes to ensure data real-time performance.

[0041] Data Classification: Collected multi-source heterogeneous data is classified according to dimensions such as data type, source device, and business-related systems. For example, server operation log data is classified according to server number, network traffic data is classified according to network device type (such as switch, router), and software application log data is classified according to application system name (such as office automation system, business transaction system).

[0042] Semantic Importance Assessment and Weighting: The semantic importance of the collected data is assessed using TF-IDF technology from Natural Language Processing. Taking server operation log data as an example, for a log entry "Server A CPU utilization reached 95% for 10 minutes," the word frequencies and inverse document frequencies of words such as "CPU utilization," "95%," and "for 10 minutes" are calculated in this log entry and the entire server operation log dataset. The TF-IDF values ​​of these words are then calculated to comprehensively evaluate the semantic importance score of the log data. According to the mapping function ,when hour, The weight corresponding to this log data .

[0043] Weighted Fusion: Assume this data processing involves three embedding models, corresponding to server operation log data, network traffic data, and software application log data, respectively, with their weight scores as follows: , , The corresponding weight , , BM25 weighted score Weight According to the weighted fusion formula Calculations yielded Based on this dataset, the operations and maintenance module continuously monitors and manages the production environment. When a certain type of data... When the value is high, it should be processed and analyzed first to ensure the continuity of critical business operations (such as business transaction systems).

[0044] II. Maintenance Module Equipment Management and Anomaly Detection Equipment Tagging and Storage: The maintenance module tags the devices (such as servers, switches, and routers) corresponding to the collected data according to the business weights obtained in step 1. For example, the server corresponding to the business transaction system is tagged based on the business weights it collects. The value is relatively high, and the business weight is marked as "high"; server data corresponding to the office automation system. The value is low, and the business weight is marked as "medium"; device data corresponding to the test environment. The lowest value is marked as "low" in terms of business weight. The marked devices are stored in different areas of the spare parts warehouse according to their weight level. High-weight devices are stored in areas near the entrance and exit of the spare parts warehouse with better environmental conditions (such as more precise temperature and humidity control) for easy access and management.

[0045] Personnel Behavior Analysis and Anomaly Detection: Real-time monitoring and data collection of personnel entering and exiting the spare parts warehouse, extracting three characteristics: frequency of visits, visit diversity, and repeat visits. A statistical period of one day was set. During this period, Personnel A entered and exited the spare parts warehouse 8 times, only active in the high-weight equipment storage area, and did not exhibit repeat visits; Personnel B entered and exited the spare parts warehouse 15 times, active in high, medium, and low-weight equipment storage areas, and repeatedly visited a certain low-weight equipment storage location 4 times.

[0046] Behavioral weight and comprehensive weight index calculation: Behavioral weight calculation rules are set based on personnel behavioral characteristics. The more visits, the wider the diversity of visits, and the lower the repeat visits, the greater the behavioral weight W(u). Calculation shows that personnel A's behavioral weight W(u) = 0.8, which corresponds to the traditional access content index. (Calculated based on data such as clicks and operation records of personnel in the spare parts warehouse management system) = 0.7; Personnel B's behavior weight W(u) = 0.3, traditional access content index According to the comprehensive weighted index formula The comprehensive weight index of person A is calculated. The comprehensive weight index of personnel B The overall weight index threshold is set at 0.3. If the overall weight index of personnel B is lower than the threshold, it is determined that there is an anomaly in their data entering and leaving the spare parts warehouse. The system automatically issues an alarm message to remind managers to further investigate personnel B's behavior.

[0047] III. Fault Prediction and Performance Evaluation of Information Technology Application Innovation Modules Failure Probability Simulation: The IT innovation module simulates the failure probability of 10 servers (IT innovation devices) corresponding to abnormal data related to personnel B's entry and exit from the spare parts warehouse, as identified by the maintenance module. Each server consists of four main components: CPU, memory, hard drive, and power supply. The failure probability of each component is obtained through statistical analysis of historical failure data. , , , According to the Monte Carlo failure probability formula Calculate the failure probability of each server. Taking one server as an example... This translates to a failure probability of approximately 6.9%. Simulations revealed that one in ten servers has a failure probability exceeding 8%, indicating a high risk of failure.

[0048] Computing power and efficiency evaluation: The computing power and efficiency of the above 10 servers were evaluated. Professional computing power testing tools were used to test the general-purpose CPU computing power and high-performance GPU computing power of each server to obtain the performance evaluation results for each server. EFLOPS EFLOPS, according to the computing power formula The computing power of the server was calculated. EFLOPS. The server's total power (PCT) was measured to be 500W using a power meter. Based on the efficiency calculation formula... The computing power of the server was calculated. FLOPS / 500W = FLOPS / W. Comparing the standard computing power (0.06 EFLOPS) and standard computing efficiency (W) of this server model. The FLOPS / W test revealed that the server's computing power and efficiency were both below the standard values. Based on the failure probability simulation results, it was determined that the server had performance issues and potential failure risks.

[0049] Anomaly Handling and Equipment Maintenance: For servers with high failure probability and low computing power and efficiency, the IT innovation module issues equipment maintenance notifications. Administrators promptly inspect and maintain the server, replacing potentially faulty components (such as hard drives) and optimizing the server's hardware configuration and software parameters. After maintenance, the server's failure probability, computing power, and efficiency are retested. The server's failure probability has decreased to 3%, computing power has increased to 0.06 EFLOPS, and efficiency has improved to [missing data]. The FLOPS / W has been restored to normal levels, ensuring the normal operation of the domestically developed equipment.

[0050] Through the implementation of this embodiment, the multi-source heterogeneous data processing efficiency of this large enterprise's domestically developed data center has been significantly improved. Key business semantic information can be quickly identified and prioritized, and the number of interruptions of high-priority businesses has been reduced by 70%. The response time for abnormal identification of spare parts warehouse data entry and exit has been shortened to less than 5 minutes, and the accuracy of abnormal identification has reached more than 90%. The failure rate of domestically developed equipment startup has been reduced from 12% to 4.5%, and the debugging time has been shortened from an average of 24 hours to 12 hours, significantly improving the business continuity and equipment reliability of the data center.

[0051] This embodiment provides a device for integrated operation and maintenance services of intelligent agent applications in the field of information technology innovation. The device is a server deployed in a large enterprise information technology innovation data center, and its hardware structure includes: Processor: It uses an Intel Xeon Gold 6348 processor, which has 28 cores and 56 threads, a base frequency of 2.6GHz, and a maximum turbo frequency of 3.9GHz. It has powerful computing capabilities and can quickly process large amounts of multi-source heterogeneous data and execute complex algorithm models (such as TF-IDF calculation model, Monte Carlo simulation algorithm, etc.).

[0052] Storage: Includes 128GB DDR4 ECC memory and a 2TB NVMe SSD solid-state drive. The DDR4 ECC memory features error checking, which improves the reliability of data storage and meets the real-time caching requirements of large amounts of data; the NVMe SSD solid-state drive has read and write speeds of over 3500MB / s, enabling rapid storage and retrieval of collected multi-source heterogeneous data, computational process data, and programs and configuration files required for system operation.

[0053] Network interface: Equipped with a dual-port 100GbE Ethernet adapter, it supports high-speed data transmission and enables high-speed data interaction with sensors, data acquisition agents, other servers, and management terminals in various computer rooms, ensuring the real-time performance and stability of data transmission.

[0054] Input / output devices include a keyboard, mouse, monitor, and USB interface, which are used by administrators to operate and configure the device, and to view system operating status, data processing results, alarm information, etc.

[0055] Power module: It adopts a 1+1 redundant power supply with an output power of 1600W, which can provide a stable power supply for the device, avoid the device shutdown due to a single power supply failure, and improve the availability of the device.

[0056] The software system of this device includes an operating system, a database management system, middleware, and the computer program corresponding to the method described in this invention. The operating system uses Linux CentOS 8.4, which offers good stability, security, and compatibility, supporting various hardware devices and software applications. The database management system uses MySQL 8.0 to store collected multi-source heterogeneous data, equipment information, personnel behavior data, fault probability data, computing power and efficiency data, etc. The middleware uses Tomcat 9.0 to deploy and run the system's web application, enabling administrators to access and operate the system through a browser. The computer program corresponding to the method described in this invention is developed using Java and built on the Spring Boot framework. It includes functional modules such as a data acquisition module, a data classification module, a semantic importance assessment module, a weight assignment and fusion module, an operation monitoring and management module, an equipment marking module, an equipment storage management module, a personnel behavior collection module, a behavior feature analysis module, a comprehensive weight index calculation and anomaly judgment module, an abnormal data association module, a fault probability simulation module, a computing power and efficiency assessment module, an abnormal startup behavior identification module, and an equipment maintenance management module. These functional modules interact and call each other through interfaces to implement the various steps of the method described in this invention.

[0057] When the device is running, the processor reads and executes the computer program from the memory. The various functional modules within the computer program work collaboratively according to the steps described in the above embodiments to achieve integrated operation and maintenance management of intelligent agent applications in the field of information technology innovation. For example, the data acquisition module receives multi-source heterogeneous data collected from various computer rooms through a network interface and stores the data in the database management system; the data classification module reads data from the database and classifies the data according to preset rules; the semantic importance assessment module assesses the semantic importance of the classified data and calculates a semantic importance score; the weight assignment and fusion module calculates weights based on the semantic importance score and merges them with BM25 weights to obtain a comprehensive weight score for the data set; the operation monitoring and management module monitors and manages the production environment based on the comprehensive weight score and issues early warning information; the maintenance module and the information technology innovation module, based on the output of the operation and maintenance module, respectively implement functions such as equipment management and anomaly detection, equipment fault prediction and performance evaluation, ultimately achieving integrated operation and maintenance services for intelligent agent applications in the field of information technology innovation.

[0058] Although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for maintaining and repairing information technology application innovation (ITAI) equipment, characterized in that, Includes the following steps: S1. Collect and classify multi-source heterogeneous data from the production environment consisting of customer data center IT systems, networks, servers, and software; use a business weight allocation algorithm to assign weights to the collected data of each embedded model based on the semantic importance of the collected data, and then merge the weight values ​​with the traditional BM25 weight values ​​to obtain a dataset. S2. Mark the devices corresponding to the collected data according to the business weights obtained in step S1, and store the marked devices in the spare parts warehouse. The behavior of personnel entering and leaving the spare parts warehouse is analyzed, and three types of features are extracted: number of visits, access diversity, and repeat visits. The behavioral weight of each person entering and leaving the warehouse in each application scenario is calculated, and then the comprehensive weight index of each person entering and leaving the warehouse is obtained. Based on the comprehensive weight index, it is determined whether the data of entering and leaving the spare parts warehouse is abnormal. S3. For the devices corresponding to the abnormal data in and out of the spare parts warehouse identified in step S2, use the Monte Carlo method to simulate the probability of device failure, combine computing power and efficiency indicators to evaluate device performance, identify abnormal device startup behavior, and ensure that the domestically developed equipment is in normal working condition.

2. The maintenance method for information technology application innovation equipment according to claim 1, characterized in that, In step S1, the semantic importance of the collected data is evaluated using natural language processing techniques, including word frequency-inverse document frequency (TF-IDF) and word vector similarity. The formula for calculating the weight assignment is: in, These are the weights of the collected data for the i-th embedded model. is the semantic importance score of the collected data of the i-th embedded model, and f is the mapping function that converts the importance score into a weight value.

3. The maintenance method for information technology application innovation equipment according to claim 1, characterized in that, The formula for calculating the weight fusion in step S1 is as follows: Where d represents the collected classification dataset, and n is the number of weights in the embedding model. and These are the weights of the i-th embedded model and the weights of BM25, respectively. and These are the weight scores of the i-th embedding model and BM25, respectively.

4. The maintenance method for information technology application innovation equipment according to claim 1, characterized in that, The formula for calculating the comprehensive weight index, i.e., the data center access content index, in step S2 is as follows: Where M represents the content accessed by the data center, U represents the set of people entering and leaving the data center, and W(u) represents the behavioral weight of the people entering and leaving the data center. The smaller the value, the greater the probability that the person entering or leaving the data center is abnormal; the larger the value, the greater the probability that the person entering or leaving the data center is a high-value user. This is the formula for calculating the content access index for people entering and leaving traditional data centers.

5. A maintenance method for information technology application innovation equipment according to claim 1, characterized in that, The formula for simulating the equipment failure probability using the Monte Carlo method in step S3 is: in, It is the failure probability of the i-th component. This indicates a product, where at least one component in the system is guaranteed to fail, meaning at least one... At that time, the failure probability of the entire system .

6. The maintenance method for information technology application innovation equipment according to claim 1, characterized in that, The formula for calculating computing power (CP) in step 3 is: in, It refers to general-purpose computing power, represented by the central processing unit (CPU). High-performance computing power, represented by graphics processing units (GPUs), is commonly measured in EFLOPS. The formula for calculating computational efficiency (CE) is: Where CP represents the computing power of the data center, expressed as a single-precision floating-point number FP32, and PCT represents the overall power of the data center IT equipment, in watts (W).

7. A maintenance system for information technology application innovation equipment, characterized in that, include: The operation and maintenance module is used to collect and classify multi-source heterogeneous data in the production environment consisting of customer data center IT systems, networks, servers, and software. It adopts a business weight allocation algorithm to assign weights to the collected data of each embedded model according to the semantic importance of the collected data, and then merges the weight values ​​with the traditional BM25 weight values ​​to obtain the data set. The maintenance module is used to mark the devices corresponding to the collected data according to the business weight obtained by the operation and maintenance module, and store the marked devices in the spare parts warehouse. The behavior of personnel entering and leaving the spare parts warehouse is analyzed, and three types of features are extracted: number of visits, access diversity, and repeat visits. The behavioral weight of each person entering and leaving the warehouse in each application scenario is calculated, and then the comprehensive weight index of each person entering and leaving the warehouse is obtained. Based on the comprehensive weight index, it is determined whether the data of entering and leaving the spare parts warehouse is abnormal. The IT innovation module is used to simulate the probability of equipment failure using the Monte Carlo method for the equipment corresponding to the abnormal data of the spare parts warehouse identified by the maintenance module, and to evaluate the equipment performance by combining computing power and efficiency indicators, and to identify abnormal equipment startup behavior.

8. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the maintenance method for domestically developed equipment as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on the electronic device, causes the electronic device to perform the maintenance method for domestically developed equipment as described in any one of claims 1 to 6.

Images