Intelligent server performance monitoring method and system

Abstract

The invention discloses a server performance intelligent monitoring method and system, and relates to the technical field of server performance, which comprises the steps of analyzing load historical data to determine the interval range and change degree of a load, providing a reliable basis for splitting the interval range of the subsequent load, considering the previous change condition of the server load, and improving the reliability of the server. Therefore, the incidence relation between the detailed interval range of the load and the performance data is accurately described. And determining an association relationship between the load and the performance historical data under different partial intervals of the load to obtain a load-performance standard association relationship, and selecting a normal value of the performance historical data from the corresponding relationship between the load and the performance historical data. According to the method, the load data and the performance data are input into the load-performance standard incidence relation to determine the deviation condition, the deviation condition is analyzed, the current performance information of the target server is output, the accuracy and adaptability of server performance monitoring are improved, and efficient and safe operation of the server is effectively guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of server performance, and in particular to a method and system for intelligently monitoring server performance. Background Art

[0002] The background technology of the intelligent server performance monitoring solution mainly stems from the continuous improvement of server stability and reliability requirements in the field of information technology. With the expansion of corporate business and the acceleration of digital transformation, the performance of servers, as the core infrastructure supporting various application systems, directly affects business continuity and user experience. Traditional server management methods often make it difficult to detect potential problems in real time and accurately, resulting in delayed fault warning and handling.

[0003] To this end, the intelligent monitoring solution came into being. It combines data collection, data analysis and early warning mechanisms. By collecting key performance indicators such as server load, response time, and throughput in real time, it uses machine learning algorithms to detect anomalies and predict performance trends. These technologies not only improve the accuracy and efficiency of monitoring, but also achieve comprehensive evaluation and optimization of server performance. The intelligent monitoring solution can promptly discover performance bottlenecks and potential failure risks, provide strong data support for the operation and maintenance team, and ensure the stable operation of the server and business continuity.

[0004] In the prior art, server performance is often monitored based on the parameter size of performance data, without considering the impact of different server loads on performance, resulting in poor accuracy and adaptability of server performance monitoring and failure to effectively ensure the safe operation of the server.

[0005] Therefore, how to improve the accuracy and adaptability of server performance monitoring is a technical problem that needs to be solved. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that the accuracy and adaptability of server performance monitoring are poor in the prior art because the influence of different load conditions is not considered, and a server performance intelligent monitoring method is proposed, which includes: Collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, split the load interval range according to the degree of change, and obtain multiple partial load intervals; Collecting historical performance data of the target server according to multiple partial intervals of the load, confirming the correlation between the load and the historical performance data in different partial intervals of the load, and obtaining a standard correlation between the load and the performance; Obtain the load data and performance data of the target server in the current period, and input the load data and performance data into the standard correlation relationship between load and performance to determine the deviation; Analyze the deviation and output the current performance information of the target server to monitor the server performance.

[0007] In some embodiments of the present application, analyzing the load history data to determine the load range and the degree of change includes: Draw a load histogram based on the load history data, mark the minimum and maximum values ​​of the load on the histogram, and use the range from the minimum to the maximum value of the load as the load interval range; Identify the load gradient section and the load jump section from the load history data, integrate the load gradient section and the load jump section respectively, and obtain the load gradient section record and the load jump section record; Based on the load gradient segment record and the load jump segment record, the occurrence frequency of different ranges under the load interval range is determined, and the frequent weights of the load gradient segment record and the load jump segment record under different ranges are determined by the occurrence frequency. Differentiate the different ranges of the load gradient segment record and the load jump segment record, and perform differential weighting in combination with the frequent weights under the corresponding range to obtain the degree of change of the gradient and the degree of change of the jump; The degree of change of the load range is determined by combining the degree of gradual change and the degree of jump change.

[0008] In some embodiments of the present application, the load interval range is split according to the degree of change to obtain multiple partial intervals of the load, including: The discrete degree of the load history data is calculated, and a split number is determined according to the discrete degree and the degree of change of the load interval range. The load interval range is correspondingly split according to the split number to obtain multiple partial intervals of the load.

[0009] In some embodiments of the present application, the correlation between the load and the historical performance data in different load intervals is confirmed to obtain a standard correlation between the load and the performance, including: Classify the performance data, determine the performance data categories corresponding to different load intervals and the distribution range of each performance data, calculate the sensitivity between the load and each performance data, and count the distribution parameters of each performance data; Distribution parameters of performance data included mean, standard deviation, quartiles, IQR, skewness, and kurtosis; The first Z score threshold was set based on the mean and standard deviation, the second Z score threshold was set based on the quartiles and IQR, and the third Z score threshold was set based on the skewness and kurtosis; The fourth Z score threshold is set by combining the first Z score threshold, the second Z score threshold, and the third Z score threshold; The fourth Z score threshold is used to distinguish normal values ​​and abnormal values ​​in the distribution range of each performance data, and the abnormal values ​​are eliminated. A standard correlation relationship between load and performance is constructed based on the normal values ​​in the distribution range of the performance data and the sensitivity between the load and each performance data.

[0010] In some embodiments of the present application, a standard association relationship between load and performance is constructed based on the normal value in the distribution range of performance data and the sensitivity between load and each performance data, including: Based on the normal value in the distribution range of performance data, the relationship between the distribution range of load and performance data in different parts of the interval is fitted to obtain a regression model. The regression coefficient in the regression model is adjusted according to the sensitivity between the load and each performance data to describe the standard correlation between load and performance.

[0011] In some embodiments of the present application, the load data and the performance data are input into a standard correlation relationship between load and performance to determine the deviation, including: Input the load data and performance data into the corresponding regression model according to different partial intervals of the load data, and count the deviations under each partial interval of the load, and draw the deviation curve of each partial interval of the load; The deviation curves of all load sections are integrated, the curve features of all deviation curves are extracted, and the deviation situation is generated.

[0012] In some embodiments of the present application, the current performance information of the target server is outputted by analyzing the deviation situation, including: Determine the curve characteristics under each type of performance data, generate the matching coefficient under each type of performance data by integrating the curve characteristics, determine the basic performance index of each type of performance data according to the parameters of each type of performance data, and determine the performance index by combining the basic performance index of each type of performance data and the matching coefficient; Integrate the performance indicators under all performance data categories to output the current performance information of the target server.

[0013] In some embodiments of the present application, the current performance information of the target server is output by integrating the performance indicators under all performance data categories, including: Integrate the performance indicators under all performance data categories to obtain the overall performance indicators of the target server; The performance indicators under all performance data categories are analyzed to obtain the performance status of each type of performance data, and the performance status, performance indicators and overall performance indicators of each type of performance data are used as the current performance information of the target server.

[0014] Correspondingly, the present application also provides a server performance intelligent monitoring system, including: The first module is used to collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, and split the load interval range according to the degree of change to obtain multiple partial load intervals; The second module is used to collect the performance history data of the target server according to multiple partial intervals of the load, confirm the correlation between the load and the performance history data in different partial intervals of the load, and obtain the standard correlation between the load and the performance; The third module is used to obtain the load data and performance data of the target server in the current period of time, and input the load data and performance data into the standard correlation relationship between load and performance to determine the deviation; The fourth module is used to analyze the deviation and output the current performance information of the target server, so as to monitor the server performance.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Analyze the load history data to determine the load interval and change degree, provide a reliable basis for the subsequent load interval splitting, take into account the previous changes in the server load, and accurately describe the relationship between the detailed load interval and performance data. Confirm the relationship between the load and performance history data in different load intervals, and obtain the standard load-performance relationship. In the corresponding relationship between the load and performance history data, select the normal value of the performance history data, and construct the standard load-performance relationship, which ensures the accuracy of the standard load-performance relationship in describing the relationship between the load and each performance data in different load intervals.

[0016] 2. Input the load data and performance data into the standard load-performance correlation to determine the deviation, analyze the deviation and output the current performance information of the target server, improve the accuracy and adaptability of server performance monitoring, and effectively ensure the efficient and safe operation of the server. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of the process of the server performance intelligent monitoring method proposed by the present invention; Figure 2 This is a schematic diagram of the structure of the server performance intelligent monitoring system proposed by the present invention. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0019] Reference Figure 1,The server performance intelligent monitoring method includes the following steps, Step S101, collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, split the load interval range according to the degree of change, and obtain multiple partial load intervals.

[0020] In this embodiment, system commands (such as uptime, top, sar, etc.) or professional monitoring tools (such as Zabbix, Nagios, Prometheus, etc.) are used to collect the load history data of the target server. Ensure that the selected tool can provide accurate and comprehensive load data, including but not limited to key indicators such as CPU usage, memory usage, network bandwidth, etc. The collected load history data is stored in a safe and reliable database or file system. Perform necessary sorting work on the data, such as deduplication, formatting, etc., to ensure the integrity and consistency of the data. The collected load history data is preprocessed, including steps such as data cleaning (such as removing outliers, missing values, etc.) and data normalization. Ensure that the preprocessed data can accurately reflect the load situation of the server. Use statistical methods (such as histograms, box plots, etc.) to analyze the preprocessed load data to determine the range of the load. In the existing means, load data is divided into three intervals: low load, medium load and high load according to its distribution. However, such vague and general load intervals are not suitable for analyzing the relationship between specific load and performance. Therefore, it is necessary to split the load intervals according to the previous changes in the server load and analyze the relationship between load and performance under specific load intervals.

[0021] In some embodiments of the present application, analyzing the load history data to determine the load range and the degree of change includes: Draw a load histogram based on the load history data, mark the minimum and maximum values ​​of the load on the histogram, and use the range from the minimum to the maximum value of the load as the load interval range; Identify the load gradient section and the load jump section from the load history data, integrate the load gradient section and the load jump section respectively, and obtain the load gradient section record and the load jump section record; Based on the load gradient segment record and the load jump segment record, the occurrence frequency of different ranges under the load interval range is determined, and the frequent weights of the load gradient segment record and the load jump segment record under different ranges are determined by the occurrence frequency. Differentiate the different ranges of the load gradient segment record and the load jump segment record, and perform differential weighting in combination with the frequent weights under the corresponding range to obtain the degree of change of the gradient and the degree of change of the jump; The degree of change of the load range is determined by combining the degree of gradual change and the degree of jump change.

[0022] In this embodiment, the load interval range is the range between the minimum and maximum values ​​of the load on the histogram, which is all possible ranges of the load. There are two modes of load change in the server, one mode is gradual change, and the other mode is direct jump. Gradual change: The load continues to increase or decrease over a period of time. This change is usually relatively stable, and it is suitable to count the frequency through interval division to reflect the overall trend of the load. Direct jump: The load changes significantly in a short period of time. This change is usually more drastic, and it is suitable to capture the load fluctuation through the statistics of specific values. All load gradient segments and all load jump segments are integrated respectively to obtain load gradient segment records and load jump segment records. Under the premise of the two modes, the frequency of occurrence of different ranges under the load interval range is calculated, and frequent weights are assigned. When calculating the difference, the difference can be weighted according to the frequency of occurrence of the load interval. Frequently occurring load intervals should be given a higher weight in the differential calculation to more accurately reflect their impact on the overall load fluctuation, and the two modes are combined to reflect the degree of load change.

[0023] In some embodiments of the present application, the load interval range is split according to the degree of change to obtain multiple partial intervals of the load, including: The discrete degree of the load history data is calculated, and a split number is determined according to the discrete degree and the degree of change of the load interval range. The load interval range is correspondingly split according to the split number to obtain multiple partial intervals of the load.

[0024] In this embodiment, the discreteness of the load history data also illustrates the load variation. The number of splits is determined by combining the discreteness and the variation of the load interval (weighted summation or other methods). The higher the discreteness and variation, the larger the number of splits, and the more accurate the analysis of the relationship between the load and performance data. Too many intervals may increase the complexity of the analysis, while too few intervals may not accurately reflect the relationship between the load and performance, so a reasonable number of splits should be determined.

[0025] Step S102, collecting the performance history data of the target server according to multiple partial intervals of the load, confirming the correlation between the load and the performance history data in different partial intervals of the load, and obtaining a standard correlation between the load and the performance.

[0026] In this embodiment, appropriate performance indicators are selected according to the application scenarios and business requirements of the server. The performance indicators should be able to fully reflect the performance status of the server, such as response time, throughput, number of concurrent users, etc. Use monitoring tools or system commands to collect performance history data according to multiple partial intervals of the load. Ensure that the collected performance data and load data are consistent in time to facilitate subsequent correlation analysis.

[0027] Server performance is reflected in multiple aspects or dimensions, each of which directly reflects the different performance conditions of the server. The following is a detailed summary and description of these aspects or dimensions: 1. CPU performance Reflection aspects: server processing speed and concurrent processing capabilities.

[0028] Performance: Processing speed: The CPU is the core component of the server and is responsible for performing various computing tasks. A high-performance CPU can perform computing tasks faster, improving the response speed and performance of the server. For example, in application scenarios that require a lot of computing, such as data analysis and scientific computing, the performance of the CPU is crucial.

[0029] Concurrent processing capability: Modern CPUs are usually multi-core designs that can handle multiple tasks at the same time. Multi-core CPUs can enhance the concurrent processing capability of the server, enabling it to handle more user requests or tasks at the same time. This is very important for running large databases, virtualized environments, or high-load applications.

[0030] 2. Memory performance Aspects: server's task processing capabilities, system stability, and database performance.

[0031] Performance: Task processing capability: Memory is used to temporarily store data generated by the CPU. Sufficient memory allows the server to process more data and requests, improving response speed. At the same time, memory also supports more programs to run simultaneously, reducing the switching time between programs and improving the system's response speed.

[0032] System stability: Larger memory capacity can alleviate the problem of system resource shortage and provide more sufficient operating space, thereby improving system stability and reliability.

[0033] Database performance: Memory, as the cache area of ​​the database, has a significant impact on the performance of the database. A larger memory capacity can provide more cache space, reduce the number of disk accesses, and significantly increase the query and write speed of the database.

[0034] 3. Disk performance Reflection aspects: server data storage and reading speed, and overall system efficiency.

[0035] Performance: Data storage and reading speed: High-performance hard disks (such as SSDs or NVMe SSDs) provide data reading and writing speeds that are much higher than traditional mechanical hard disks (HDDs). This is directly related to the response time of applications and the speed of data processing. For example, operations such as database queries, logging, and file transfers can be significantly accelerated.

[0036] Overall system efficiency: Faster disk read and write speeds mean shorter startup times for the operating system and applications, and higher overall system efficiency. At the same time, high-performance hard drives can also support higher IOPS (input and output operations per second), ensuring stable operation under high load.

[0037] 4. Network performance Reflection aspects: server's data transmission speed and concurrent request processing capabilities.

[0038] Performance: Data transmission speed: Network bandwidth determines the speed at which data can be transmitted in the network. High network bandwidth can reduce data transmission delays and improve the efficiency of distributed computing and remote access.

[0039] Concurrent request processing capability: In high-concurrency scenarios, the server needs to process a large number of requests at the same time. Sufficient network bandwidth can support more concurrent connections and ensure that the server can process these requests in a timely manner.

[0040] 5. Reliability and stability Reflection aspects: server's fault tolerance and long-term stable operation capabilities.

[0041] Performance: Fault tolerance: High-quality server hardware and software design can improve the server's fault tolerance. For example, redundant power supplies and fans, predictable hard disk and fan failures, and RAID systems are all common technologies used to improve server reliability.

[0042] Long-term stable operation capability: The stability and reliability of a server are also reflected in its ability to run stably for a long time without failure or performance degradation. This is crucial for application scenarios that require continuous service provision (such as websites, databases, etc.).

[0043] In this embodiment, the performance history data of the target server is collected according to multiple partial intervals of the load, and the performance history data of the target server corresponding to each interval is collected based on the partial intervals of the load.

[0044] In some embodiments of the present application, the correlation between the load and the historical performance data in different load intervals is confirmed to obtain a standard correlation between the load and the performance, including: Classify the performance data, determine the performance data categories corresponding to different load intervals and the distribution range of each performance data, calculate the sensitivity between the load and each performance data, and count the distribution parameters of each performance data; Distribution parameters of performance data included mean, standard deviation, quartiles, IQR, skewness, and kurtosis; The first Z score threshold was set based on the mean and standard deviation, the second Z score threshold was set based on the quartiles and IQR, and the third Z score threshold was set based on the skewness and kurtosis; The fourth Z score threshold is set by combining the first Z score threshold, the second Z score threshold, and the third Z score threshold; The fourth Z score threshold is used to distinguish normal values ​​and abnormal values ​​in the distribution range of each performance data, and the abnormal values ​​are eliminated. A standard correlation relationship between load and performance is constructed based on the normal values ​​in the distribution range of the performance data and the sensitivity between the load and each performance data.

[0045] In this embodiment, the distribution range of each performance data (the distribution range of the parameter value) is used to calculate the sensitivity between the load and each performance data. Here, the sensitivity can be calculated by methods such as the Pearson correlation coefficient. In order to accurately describe the relationship between load and performance, the abnormal data (error, abnormal performance, etc.) in the performance data are removed. Here, the Z-score method is used to identify normal data and abnormal data, calculate the Z-score of each data, and compare the Z-score threshold to determine whether it is abnormal data.

[0046] In this embodiment, the distribution parameters of the performance data include mean, standard deviation, quartile, IQR, skewness and kurtosis, which are as follows: Mean: The mean is the most common method that helps us understand the central tendency of the data. In the Z-score method, the mean is used to calculate the difference between each data point and the mean.

[0047] Median: The median is another indicator of the central tendency of a data set. It is the value in the middle after the data is arranged in order of size. The median is not affected by extreme values. When there are extreme values ​​in the data set, the median can more accurately reflect the central position of the data than the mean.

[0048] Mode: The mode is the value that occurs most often in a data set. When analyzing the distribution of data, the mode can help us understand the most common situation.

[0049] Quartiles: Quartiles can divide data into four parts, each containing 25% of the data points. The distance between the first quartile (Q1) and the third quartile (Q3) is called the interquartile range (IQR), which is an indicator of the degree of dispersion of data. Quartiles can help us quickly identify outliers, and data points that fall outside Q1-1.5IQR and Q3+1.5IQR are usually considered outliers.

[0050] Skewness: Skewness describes the symmetry of the data distribution. A positive skewness indicates that the data distribution is skewed to the right, i.e., there are more large values; a negative skewness indicates that the data distribution is skewed to the left, i.e., there are more small values. Skewness can affect the setting of the Z-score threshold, because a data distribution with a large skewness may be more likely to produce extreme values.

[0051] Kurtosis: Kurtosis describes the peakedness of the data distribution. High kurtosis indicates that the data is concentrated around the mean, while low kurtosis indicates that the data is more dispersed. Kurtosis can also affect the setting of the Z-score threshold, because data distributions with high kurtosis may be more likely to produce data points close to the mean.

[0052] Based on mean and standard deviation: When the data is close to normal distribution, it is usually possible to use the mean and standard deviation to calculate the Z-score and set a fixed threshold (such as 3) to identify outliers. However, if the data distribution deviates from normal distribution or is more volatile, you may need to adjust the threshold.

[0053] Consider quartiles and IQR: For data with high volatility or data that deviates from a normal distribution, you can use quartiles and IQR to assist in setting the Z-score threshold. For example, you can calculate the IQR of the data and use Q1-1.5IQR and Q3+1.5IQR as the boundaries of outliers. Then, adjust the Z-score threshold based on these boundaries.

[0054] Combine skewness and kurtosis: When considering setting a Z-score threshold, you can also consider the skewness and kurtosis of the data. If the data is more skewed or has higher kurtosis, you may need to set a more relaxed threshold to avoid misclassifying normal fluctuations as abnormal.

[0055] In this embodiment, the fourth Z score threshold is set by combining the first Z score threshold, the second Z score threshold, and the third Z score threshold. The calculation formula is as follows: ; in, The load The corresponding The fourth Z-score threshold for performance data, , , are the contribution weights of the first Z score threshold, the second Z score threshold, and the third Z score threshold, respectively. , , The load The corresponding The first Z score threshold, the second Z score threshold, and the third Z score threshold of the performance data are and They are , , The maximum and minimum of the three, The load The corresponding The first constant of a performance data; In this embodiment, It represents the correction of the average of the maximum and minimum values ​​of the influence of the first Z score threshold, the second Z score threshold, and the third Z score threshold to the average of the sum of the three. In order to balance the size of the correction function.

[0056] In some embodiments of the present application, a standard association relationship between load and performance is constructed based on the normal value in the distribution range of performance data and the sensitivity between load and each performance data, including: Based on the normal value in the distribution range of performance data, the relationship between the distribution range of load and performance data in different parts of the interval is fitted to obtain a regression model. The regression coefficient in the regression model is adjusted according to the sensitivity between the load and each performance data to describe the standard correlation between load and performance.

[0057] In this embodiment, based on the normal value in the distribution range of the performance data, a suitable regression model (such as linear regression, polynomial regression, etc.) is selected to fit the relationship between the distribution range of the load and the performance data in different partial intervals. The model is trained using the collected data to obtain a preliminary regression coefficient. The sensitivity coefficient reflects the degree of influence of the load change on the performance data change. If the sensitivity coefficient is large, it means that the load has a more significant impact on the performance data; otherwise, the impact is small. According to the results of the sensitivity analysis, the regression coefficient in the regression model is adjusted. If a certain performance data is highly sensitive to the load, the regression coefficient corresponding to the performance data can be appropriately increased to enhance the impact of the load on the performance data. On the contrary, if a certain performance data is less sensitive to the load, the regression coefficient corresponding to the performance data can be appropriately reduced to weaken the impact of the load on the performance data. The adjusted regression model is verified using a validation data set to evaluate the prediction accuracy and stability of the model. If the model performs poorly, the above steps can be repeated to optimize the model until a satisfactory model is obtained, and the standard correlation relationship between load and performance in different load intervals is described by different regression models.

[0058] Step S103, obtaining the load data and performance data of the target server in the current period, and inputting the load data and performance data into the standard association relationship between load and performance to determine the deviation.

[0059] In this embodiment, real-time data collection: use monitoring tools to obtain the load data and performance data of the target server in real time within a period of time. Data recording: record the acquired data and store it in a safe and reliable database or file system. The data record should include two parts: load data and performance data, so as to facilitate subsequent correlation analysis and monitoring. The load data and performance data are input into the corresponding regression model to count the deviation of the performance data.

[0060] In some embodiments of the present application, the load data and the performance data are input into a standard correlation relationship between load and performance to determine the deviation, including: Input the load data and performance data into the corresponding regression model according to different partial intervals of the load data, and count the deviations under each partial interval of the load, and draw the deviation curve of each partial interval of the load; The deviation curves of all load sections are integrated, the curve features of all deviation curves are extracted, and the deviation situation is generated.

[0061] In this embodiment, the performance deviation is the difference between the predicted value and the actual value corresponding to different loads under the regression model. The smaller the deviation, the better the match between the actual load and the performance. It also describes the match between the actual load-performance relationship and the standard load-performance relationship. The curve characteristics of the deviation curve include slope, extreme value, smoothness, etc.

[0062] Step S104, analyzing the deviation and outputting the current performance information of the target server, thereby realizing monitoring of the server performance.

[0063] In this embodiment, the current performance of the target server is analyzed in combination with the parameter size and deviation of the performance data, and the current performance is comprehensively analyzed from the matching of the parameter size of the performance data and the load-performance relationship.

[0064] In some embodiments of the present application, the current performance information of the target server is outputted by analyzing the deviation situation, including: Determine the curve characteristics under each type of performance data, generate the matching coefficient under each type of performance data by integrating the curve characteristics, determine the basic performance index of each type of performance data according to the parameters of each type of performance data, and determine the performance index by combining the basic performance index of each type of performance data and the matching coefficient; Integrate the performance indicators under all performance data categories to output the current performance information of the target server.

[0065] In some embodiments of the present application, the current performance information of the target server is output by integrating the performance indicators under all performance data categories, including: Integrate the performance indicators under all performance data categories to obtain the overall performance indicators of the target server; The performance indicators under all performance data categories are analyzed to obtain the performance status of each type of performance data, and the performance status, performance indicators and overall performance indicators of each type of performance data are used as the current performance information of the target server.

[0066] In this embodiment, all curve features are integrated (weighted sum after normalization, and mapped to obtain the matching coefficient) to obtain the matching coefficient between this type of performance data and the load, and the parameter size of the indicator data of each type of performance is evaluated to obtain the basic performance indicator (the quality of the performance parameters). The performance indicators under all performance data categories are integrated to output the current performance information of the target server.

[0067] ; in, is the current performance information of the target server (total server performance), is the number of performance data categories, For the The matching coefficient of performance data, For the The basic performance indicators of performance data, for The number of performance data categories that are less than the corresponding threshold, is the performance difference conversion factor, for The value less than the corresponding threshold The matching coefficient of performance data, for The value less than the corresponding threshold The basic performance indicators of performance data, is the preset constant, express The correction of the sum of the performance of all performance data that is less than the sum of the corresponding threshold, This is to balance the correction function.

[0068] Correspondingly, this application also provides a server performance intelligent monitoring system, such as Figure 2 As shown, including The first module is used to collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, and split the load interval range according to the degree of change to obtain multiple partial load intervals; The second module is used to collect the performance history data of the target server according to multiple partial intervals of the load, confirm the correlation between the load and the performance history data in different partial intervals of the load, and obtain the standard correlation between the load and the performance; The third module is used to obtain the load data and performance data of the target server in the current period of time, and input the load data and performance data into the standard correlation relationship between load and performance to determine the deviation; The fourth module is used to analyze the deviation and output the current performance information of the target server, so as to monitor the server performance.

[0069] Compared with the prior art, the present invention has the following beneficial effects: 1. Analyze the load history data to determine the load interval and change degree, provide a reliable basis for the subsequent load interval splitting, take into account the previous changes in the server load, and accurately describe the relationship between the detailed load interval and performance data. Confirm the relationship between the load and performance history data in different load intervals, and obtain the standard load-performance relationship. In the corresponding relationship between the load and performance history data, select the normal value of the performance history data, and construct the standard load-performance relationship, which ensures the accuracy of the standard load-performance relationship in describing the relationship between the load and each performance data in different load intervals.

[0070] 2. Input the load data and performance data into the standard load-performance correlation to determine the deviation, analyze the deviation and output the current performance information of the target server, improve the accuracy and adaptability of server performance monitoring, and effectively ensure the efficient and safe operation of the server.

[0071] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0072] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present invention.

[0073] Those skilled in the art will appreciate that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0074] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for intelligently monitoring server performance, characterized in that: include, Collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, split the load interval range according to the degree of change, and obtain multiple partial load intervals; Collecting historical performance data of the target server according to multiple partial intervals of the load, confirming the correlation between the load and the historical performance data in different partial intervals of the load, and obtaining a standard correlation between the load and the performance; Obtain the load data and performance data of the target server in the current period, and input the load data and performance data into the standard correlation relationship between load and performance to determine the deviation; Analyze the deviation and output the current performance information of the target server to monitor the server performance.

2. The server performance intelligent monitoring method according to claim 1, characterized in that: Analyze historical load data to determine the load range and degree of change, including: Draw a load histogram based on the load history data, mark the minimum and maximum values ​​of the load on the histogram, and use the range from the minimum to the maximum value of the load as the load interval range; Identify the load gradient section and the load jump section from the load history data, integrate the load gradient section and the load jump section respectively, and obtain the load gradient section record and the load jump section record; Based on the load gradient segment record and the load jump segment record, the occurrence frequency of different ranges under the load interval range is determined, and the frequent weights of the load gradient segment record and the load jump segment record under different ranges are determined by the occurrence frequency. Differentiate the different ranges of the load gradient segment record and the load jump segment record, and perform differential weighting in combination with the frequent weights under the corresponding range to obtain the degree of change of the gradient and the degree of change of the jump; The degree of change of the load range is determined by combining the degree of gradual change and the degree of jump change.

3. The server performance intelligent monitoring method according to claim 2, characterized in that: The load interval range is split according to the degree of change to obtain multiple partial intervals of the load. include, The discrete degree of the load history data is calculated, and a split number is determined according to the discrete degree and the degree of change of the load interval range. The load interval range is correspondingly split according to the split number to obtain multiple partial intervals of the load.

4. The server performance intelligent monitoring method according to claim 1, characterized in that: Confirm the correlation between load and performance historical data in different load intervals, and obtain the standard correlation between load and performance, including: Classify the performance data, determine the performance data categories corresponding to different load intervals and the distribution range of each performance data, calculate the sensitivity between the load and each performance data, and count the distribution parameters of each performance data; Distribution parameters of performance data included mean, standard deviation, quartiles, IQR, skewness, and kurtosis; The first Z score threshold was set based on the mean and standard deviation, the second Z score threshold was set based on the quartiles and IQR, and the third Z score threshold was set based on the skewness and kurtosis; The fourth Z score threshold is set by combining the first Z score threshold, the second Z score threshold, and the third Z score threshold; The fourth Z score threshold is used to distinguish normal values ​​and abnormal values ​​in the distribution range of each performance data, and the abnormal values ​​are eliminated. A standard correlation relationship between load and performance is constructed based on the normal values ​​in the distribution range of the performance data and the sensitivity between the load and each performance data.

5. The server performance intelligent monitoring method according to claim 4 is characterized in that: Based on the normal value in the distribution range of performance data and the sensitivity between load and each performance data, a standard load-performance correlation is constructed, including: Based on the normal value in the distribution range of performance data, the relationship between the distribution range of load and performance data in different parts of the interval is fitted to obtain a regression model. The regression coefficient in the regression model is adjusted according to the sensitivity between the load and each performance data to describe the standard correlation between load and performance.

6. The server performance intelligent monitoring method according to claim 5, characterized in that: The load and performance data are entered into the standard load-performance correlation to determine deviations, including, Input the load data and performance data into the corresponding regression model according to different partial intervals of the load data, and count the deviations under each partial interval of the load, and draw the deviation curve of each partial interval of the load; The deviation curves of all load sections are integrated, the curve features of all deviation curves are extracted, and the deviation situation is generated.

7. The method for intelligently monitoring server performance according to claim 6, characterized in that: Analyze deviations and output current performance information of the target server, including: Determine the curve characteristics under each type of performance data, generate the matching coefficient under each type of performance data by integrating the curve characteristics, determine the basic performance index of each type of performance data according to the parameters of each type of performance data, and determine the performance index by combining the basic performance index of each type of performance data and the matching coefficient; Integrate the performance indicators under all performance data categories to output the current performance information of the target server.

8. The method for intelligently monitoring server performance according to claim 7, characterized in that: Integrate the performance indicators under all performance data categories to output the current performance information of the target server. include, Integrate the performance indicators under all performance data categories to obtain the overall performance indicators of the target server; The performance indicators under all performance data categories are analyzed to obtain the performance status of each type of performance data, and the performance status, performance indicators and overall performance indicators of each type of performance data are used as the current performance information of the target server.

9. Server performance intelligent monitoring system, characterized in that: include, The first module is used to collect the load history data of the target server, analyze the load history data to determine the load interval range and the degree of change, and split the load interval range according to the degree of change to obtain multiple partial load intervals; The second module is used to collect the performance history data of the target server according to multiple partial intervals of the load, confirm the correlation between the load and the performance history data in different partial intervals of the load, and obtain the standard correlation between the load and the performance; The third module is used to obtain the load data and performance data of the target server in the current period of time, and input the load data and performance data into the standard correlation relationship between load and performance to determine the deviation; The fourth module is used to analyze the deviation and output the current performance information of the target server, so as to monitor the server performance.

Images