Performance capacity model construction methods, apparatus, equipment and computer program products

Abstract

This application relates to the field of performance capacity management technology, and provides a method, apparatus, device, and computer program product for constructing a performance capacity model. The method includes: synthesizing corresponding sample data based on the correlation between business volume data from various interface channels and their corresponding first CPU usage data; analyzing the sample data using a preset principal component regression analysis method to obtain a corresponding comprehensive variable equation and its corresponding comprehensive variables; training the comprehensive variable equation using the comprehensive variables as independent variables and second CPU usage data to construct a corresponding performance capacity model. The performance capacity model construction method provided in this application does not require manual intervention; it automatically analyzes the sample data using principal component regression analysis to construct an accurate performance capacity model. This accurate performance capacity model can accurately estimate the performance capacity of each product, improving the accuracy of performance capacity estimation.

Description

Technical Field

[0001] This application relates to the field of performance and capacity management technology, and in particular to a method, apparatus, device, and computer program product for constructing a performance and capacity model. Background Technology

[0002] In traditional IT (Internet Technology) infrastructure environments, the main methods for performance capacity estimation include: performance capacity planning through monitoring systems, performance capacity planning through stress testing, and performance capacity estimation through human experience.

[0003] Performance capacity planning methods for monitoring systems: Set performance alarm thresholds in the monitoring system and monitor the trend of these thresholds using a performance trend chart. If the performance alarm threshold is reached, notify operations personnel to expand capacity. Performance capacity planning methods for stress testing: Before system deployment, manually build an offline test environment and conduct offline tests. Determine the system's limits through stress testing and by combining the results with the monitoring system data. Performance capacity estimation methods based on human experience: Combine human operations experience with the maximum traffic within a certain time period as a short-term traffic estimate.

[0004] In summary, the current performance capacity estimation methods all adopt the "human experience + stress test" approach, that is, they judge whether expansion is needed based on past human operation and maintenance experience and current system monitoring information. Since different users have different operation and maintenance experience, the accuracy of performance capacity estimation is low. Summary of the Invention

[0005] This application provides a method, apparatus, device, and computer program product for constructing a performance capacity model, which aims to improve the accuracy of performance capacity estimation.

[0006] Firstly, this application provides a method for constructing a performance capacity model, including:

[0007] Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, corresponding sample data are synthesized.

[0008] The sample data are analyzed according to the preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0009] The comprehensive variable equation is trained using the comprehensive variable as the independent variable data and the second CPU usage data to construct the corresponding performance capacity model.

[0010] In one embodiment, the step of analyzing the sample data according to a preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables includes:

[0011] After removing the CPU usage data consumed by the background monitoring business, fault handling business and upgrade and maintenance business in the sample data, the principal component data corresponding to the sample data is obtained;

[0012] Dimensionality reduction analysis is performed on the principal component data based on a preset analysis matrix to obtain the target principal component data and its corresponding component matrix.

[0013] Based on the target principal component data and the component matrix, the comprehensive variable equation and its corresponding comprehensive variables are determined.

[0014] The step of determining the comprehensive variable equation and its corresponding comprehensive variables based on the target principal component data and the component matrix includes:

[0015] The business volume data in the component matrix is ​​multiplied one-to-one with the CPU usage data corresponding to the target principal component data in the component matrix to construct the corresponding calculation equation;

[0016] The calculation equation is determined as the comprehensive variable equation, and the equation result corresponding to the comprehensive variable equation is determined as the comprehensive variable.

[0017] The steps of training the comprehensive variable equation using comprehensive variable data and second CPU usage data to construct the corresponding performance capacity model include:

[0018] Regression analysis was performed using the composite variable as the independent variable and the second CPU data to obtain the corresponding regression results;

[0019] The regression results are used to train the comprehensive variable equation to construct the corresponding principal component regression model, and the performance capacity model is constructed based on the principal component regression model.

[0020] After the step of training the comprehensive variable equation with comprehensive variable data and second CPU usage data to construct the corresponding performance capacity model, the method further includes:

[0021] Based on the performance capacity model and the data relationship between the service ingress traffic data and the cluster traffic data of each cluster, the estimated traffic data of each cluster is estimated.

[0022] Based on the estimated traffic data of each cluster and the corresponding number of machines, estimate the module traffic data of each module in each cluster.

[0023] Based on the preset module calculation formula and the module traffic data of each module in each cluster, estimate the module CPU usage data of each module in each cluster.

[0024] The overall system capacity is estimated based on the CPU usage data of each module in each cluster.

[0025] The step of estimating the overall system capacity based on the module CPU usage data of each module in each cluster includes:

[0026] The first single-machine CPU usage data is obtained based on the process CPU usage data of the target machine and the sum of the module CPU usage data of all modules in the target machine.

[0027] Based on the process CPU usage data and the module CPU usage data of any module of the target machine, the second single-machine CPU usage data is obtained.

[0028] The overall capacity of the machine is estimated based on the first single-machine CPU usage data and the second single-machine CPU usage data.

[0029] The feature is that, after the step of estimating the overall machine capacity based on the module CPU usage data of each module in each of the clusters, it further includes:

[0030] Based on the access traffic data of the newly added service volume data and the overall machine capacity, the resource usage data of the newly added service volume data is determined.

[0031] Secondly, this application also provides a performance capacity model construction apparatus comprising:

[0032] The synthesis module is used to synthesize corresponding sample data based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data;

[0033] The analysis module is used to analyze the sample data according to a preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0034] The module is used to train the equation of the comprehensive variable with the comprehensive variable as the independent variable data and the data used by the second CPU, and to build the corresponding performance capacity model.

[0035] Thirdly, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the performance capacity model construction method described in the first aspect.

[0036] Fourthly, this application also provides a computer program product, the computer program product comprising a computer program, which, when executed by the processor, implements the steps of the performance capacity model construction method described in the first aspect.

[0037] The performance capacity model construction method, apparatus, equipment, and computer program product provided in this application require no manual intervention during the performance capacity model construction process. They automatically analyze sample data using principal component regression analysis to construct an accurate performance capacity model. This accurate model can accurately estimate the performance capacity of each product, improving the accuracy of performance capacity estimation. Furthermore, during the construction of the performance capacity model, the principal component regression analysis method determines a comprehensive variable equation, ensuring that the equation fully represents the relationship between business volume data and CPU usage data. This makes the constructed performance capacity model more scientific and further improves the accuracy of performance capacity estimation. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is one of the flowcharts illustrating the performance capacity model construction method provided in this application;

[0040] Figure 2 This is the second flowchart illustrating the performance capacity model construction method provided in this application;

[0041] Figure 3 This is a schematic diagram of the performance capacity model building device provided in this application;

[0042] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] The following is combined Figures 1 to 4This application describes the performance capacity model construction method, apparatus, device, and computer program product provided.

[0045] Specifically, this application provides a method for constructing a performance capacity model, referring to... Figure 1 , Figure 1 This is one of the flowcharts illustrating the performance capacity model construction method provided in this application.

[0046] This application provides an embodiment of a performance capacity model construction method. It should be noted that although the logical order is shown in the flowchart, under certain data conditions, the steps shown or described may be completed in a different order than that shown here.

[0047] This application uses an electronic device as the execution subject for example, and a management system as one of the manifestations of the electronic device, but does not limit the electronic device.

[0048] The performance capacity model construction method provided in this application includes:

[0049] Step S10: Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, synthesize the corresponding sample data.

[0050] It should be noted that when the management system receives a model building instruction from the user terminal, it builds a performance capacity model based on the instruction. Alternatively, the management system may automatically trigger the building of the performance capacity model at regular intervals; the latter is used as an example in this embodiment.

[0051] When a management system needs to build a performance capacity model, it needs to collect business access volume data (business volume data) and system performance data. System performance data is structured data and can be obtained from existing monitoring systems using SQL (Structured Query Language) statements. Business volume data needs to be collected based on the characteristics of the business system, which includes, but is not limited to, ERP (Enterprise Resource Planning), OA (Office Automation), and CRM (Customer Relationship Management). This embodiment uses NGCRM (Next Generation Customer Relationship Management) as an example. This embodiment collects NGCRM business volume data through interface call methods and order log methods. It should be noted that the business volume data collection methods in this embodiment (interface call method and order log method) can directly obtain the atomic service data of NGCRM business requests without data mapping or transformation, resulting in highly accurate business volume data. Next, the management system stores the business volume data and system performance data in a database. It then uses SQL statements to aggregate and perform correlation analysis on the data, obtaining the business volume data and corresponding CPU (Central Processing Unit) usage data for each interface channel. CPU usage data is one form of system performance data and can be understood as CPU utilization rate. Finally, based on the correlation between the business volume data and the corresponding first CPU usage data for each interface channel, the management system synthesizes the business volume data and their corresponding first CPU usage data for each interface channel to obtain the corresponding sample data.In this embodiment, the order volume data of a certain NGCRM business system includes front-end channel business volume data, online hall channel business volume data, customer service channel business volume data, short hall channel business volume data, mobile hall channel business volume data, batch channel business volume data, back-end channel business volume data, and other channel business volume data, with corresponding CPU usage data of 0.575, 0.729, 0.814, 0.801, 0.632, 0.186, 0.095, and 0.591, respectively; the interface call business volume data includes online hall channel business volume data, customer service channel business volume data, short hall channel business volume data, mobile hall channel business volume data, and other channel business volume data, with corresponding CPU usage data of 0.700, 0.937, 0.832, 0.581, and 0.738, respectively. The sample data of this certain NGCRM business system can be represented as shown in Table 1, which is the sample data.

[0052] Table 1 Sample Data

[0053]

[0054] Step S20: Analyze the sample data according to the preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables.

[0055] It should be noted that the steps of the preset component regression analysis method in this embodiment include, but are not limited to, sample data cleaning, principal component data dimensionality reduction, and regression analysis.

[0056] The management system performs sample data cleaning, specifically removing CPU usage data from abnormal service operations. Abnormal service operations include, but are not limited to, backend monitoring, fault handling, and upgrade / maintenance services. This removes CPU usage data from normal service operations, which is the principal component data of the sample data. Next, the management system performs dimensionality reduction analysis on the principal component data to obtain principal component data and its corresponding component matrix that represent all variable information; this is the target principal component data and its corresponding component matrix. Finally, based on the target principal component data and its corresponding component matrix, the management system constructs various formulas relating service volume data and CPU usage data, i.e., various comprehensive variable equations for the sample data. Simultaneously, the management system determines the corresponding comprehensive variables based on these comprehensive variable equations, as described in steps S201 to S203.

[0057] Furthermore, the specific descriptions of steps S201 to S203 are as follows:

[0058] Step S201: Remove the CPU usage data consumed by the background monitoring business, fault handling business and upgrade and maintenance business in the sample data to obtain the principal component data corresponding to the sample data;

[0059] Step S202: Perform dimensionality reduction analysis on the principal component data based on the preset analysis matrix to obtain the target principal component data and its corresponding component matrix;

[0060] Step S203: Determine the comprehensive variable equation and its corresponding comprehensive variables based on the target principal component data and the component matrix.

[0061] The management system removes CPU usage data from the sample data for background monitoring, fault handling, and upgrade / maintenance operations, obtaining principal component data corresponding to the sample data. In other words, this embodiment only retains principal component data that reflects most business scenarios or normal business conditions, resulting in higher accuracy and stronger representativeness. The performance capacity model trained using this principal component data is more scientific and accurate. Next, the management system obtains a preset analysis matrix, where the covariance matrix is ​​one form of the preset analysis matrix. Table 2 shows one form of the preset covariance matrix.

[0062] Table 2 Preset covariance matrix

[0063]

[0064] Next, the management system performs dimensionality reduction analysis on the principal component data based on the preset variance value in the preset covariance matrix to obtain the target principal component data. In this embodiment, if the preset variance value is 97.1%, it can be understood that after dimensionality reduction analysis of the principal component data, the number of target principal component data obtained is 2. The two target principal component data can be defined as component 1 and component 2, respectively. Furthermore, the CPU usage data of component 1 and component 2 corresponding to the business volume data of each channel can be obtained based on the variance value and total value in the preset covariance matrix.

[0065] Specifically, the management system determines the contribution of each channel's business volume data to the total value in the variance value, and assigns this contribution value as the corresponding CPU usage data. In this embodiment, the variance value corresponding to a difference of 97.1% is 16.186 and the total value is 34464737244.918. The contribution values ​​of the front-end channel's business volume data to the total value in the variance value are 0.609 and 0.033; the online hall channel's business volume data's contribution values ​​are 0.870 and 0.494; the customer service channel's business volume data's contribution values ​​are 0.938 and 0.342; the short-term hall channel's business volume data's contribution values ​​are 0.859 and 0.226; the mobile hall channel's business volume data's contribution values ​​are 0.653 and 0.058; and the other channel's business volume data's contribution values ​​are 0.662 and 0.212. The component matrix can then be shown in Table 3, which is the component matrix.

[0066] Table 3 Component Matrix

[0067] Business volume data Ingredients 1 Ingredient 2 Front-end channel business volume data 0.609 0.033 Online hall channel business volume data 0.870 0.494 Customer service channel business volume data 0.938 0.342 Short-terminal channel business volume data 0.859 0.226 Business volume data of the mobile terminal channel 0.653 0.058 Business volume data from other channels 0.662 0.212

[0068] Finally, the management system constructs corresponding comprehensive variable equations based on the target principal component data and its corresponding component matrix, and then determines the corresponding comprehensive variables based on the comprehensive variable equations. The specific steps are as described in steps S2031 to S2032.

[0069] This embodiment uses principal component regression analysis to determine the comprehensive variable equation, which can comprehensively represent the relationship between business volume data and CPU usage data, thus making the comprehensive variable equation more scientific.

[0070] Furthermore, the specific descriptions of steps S2031 to S2032 are as follows:

[0071] Step S2031: Multiply the business volume data in the component matrix with the CPU usage data corresponding to the target principal component data in the component matrix one by one to construct the corresponding calculation equation;

[0072] Step S2032: Determine the calculation equation as the comprehensive variable equation, and determine the equation result corresponding to the comprehensive variable equation as the comprehensive variable.

[0073] Specifically, the management system multiplies each business volume data in the component matrix with the CPU usage data corresponding to the target principal component data in the component matrix one-to-one, constructing corresponding calculation equations. The number of calculation equations corresponds to the number of target principal component data. In the above embodiment, the number of target principal component data is 2, therefore the number of calculation equations is 2. The two calculation equations are as follows: Calculation Equation 1: Front-end channel business volume data * 0.609 + Online hall channel business volume data * 0.870 + Customer service channel business volume data * 0.938 + Short hall channel business volume data * 0.859 + Mobile hall channel business volume data * 0.653 + Other channel business volume data * 0.662; Calculation Equation 2: Front-end channel business volume data * 0.033 + Online hall channel business volume data * 0.494 + Customer service channel business volume data * 0.342 + Short hall channel business volume data * 0.226 + Mobile hall channel business volume data * 0.058 + Other channel business volume data * 0.212.

[0074] After obtaining the calculation equations, the management system will determine them as comprehensive variable equations, namely: Comprehensive Variable Equation 1: Front-end channel business volume data * 0.609 + Online hall channel business volume data * 0.870 + Customer service channel business volume data * 0.938 + Short video hall channel business volume data * 0.859 + Mobile hall channel business volume data * 0.653 + Other channel business volume data * 0.662; Comprehensive Variable Equation 2: Front-end channel business volume data * 0.033 + Online hall channel business volume data * 0.494 + Customer service channel business volume data * 0.342 + Short video hall channel business volume data * 0.226 + Mobile hall channel business volume data * 0.058 + Other channel business volume data * 0.212.

[0075] Finally, the management system determines the equation results corresponding to the comprehensive variable equations as comprehensive variables, namely: Comprehensive Variable 1 = Front-end channel business volume data * 0.609 + Online hall channel business volume data * 0.870 + Customer service channel business volume data * 0.938 + Short video hall channel business volume data * 0.859 + Mobile hall channel business volume data * 0.653 + Other channel business volume data * 0.662; Comprehensive Variable 2 = Front-end channel business volume data * 0.033 + Online hall channel business volume data * 0.494 + Customer service channel business volume data * 0.342 + Short video hall channel business volume data * 0.226 + Mobile hall channel business volume data * 0.058 + Other channel business volume data * 0.212.

[0076] This embodiment constructs a comprehensive variable equation by using the business volume data in the component matrix and the CPU usage data corresponding to the target principal component data in the component matrix. This comprehensive variable equation can fully represent the relationship between business volume data and CPU usage data, thus making the comprehensive variable equation more scientific.

[0077] Step S30: Using the comprehensive variable as the independent variable data and the second CPU usage data, train the comprehensive variable equation to construct the corresponding performance capacity model.

[0078] The management system uses some of the comprehensive variables as independent variable data and some of the comprehensive variables as CPU usage data corresponding to their independent variable data. The comprehensive variable equation is trained by using the independent variable data and their corresponding CPU usage data to construct the corresponding performance capacity model, as described in steps S301 to S302.

[0079] Furthermore, the specific descriptions of steps S301 to S302 are as follows:

[0080] Step S301: Perform regression analysis using the comprehensive variable as the independent variable data and the second CPU data to obtain the corresponding regression results;

[0081] Step S302: Train the comprehensive variable equation using the regression results to construct the corresponding principal component regression model, and construct the performance capacity model based on the principal component regression model.

[0082] Specifically, the management system performs regression analysis on each comprehensive variable as independent variable data and its corresponding second CPU usage data to obtain the corresponding regression results. Then, the management system substitutes the regression results into the comprehensive variable equation and trains the comprehensive variable equation using the regression results to construct the corresponding principal component regression model. In this embodiment, the number of comprehensive variables is two: one comprehensive variable is the independent variable data, and the other is its corresponding CPU usage data. For ease of explanation, comprehensive variable 1 is used as the independent variable data, and comprehensive variable 2 is the CPU usage data. After regression analysis and substitution into the comprehensive variable equation, the corresponding regression coefficients, t-values ​​(significance test values), significance, model summative R-values ​​(correlation coefficient values), and standard error values ​​are obtained. Based on the regression coefficients, t-values, significance, model summative R-values, and standard error values, the corresponding principal component regression model is constructed.

[0083] Furthermore, taking the aforementioned business volume data as an example, for the front-end channel business volume data, the independent variable is X1, and the CPU usage data is Y1. After regression analysis and substituting into the comprehensive variable equation, the regression coefficient is 1.75E-05 (1.75 to the power of -5). For the online hall channel business volume data, the independent variable is X2, and the CPU usage data is Y2. After regression analysis and substituting into the comprehensive variable equation, the regression coefficient is 2.50E-05. For the customer service channel business volume data, the independent variable is X3, and the CPU usage data is Y3. After regression analysis and substituting into the comprehensive variable equation, ... The regression coefficient was 2.70E-05. For the short-term hall channel business volume data, the independent variable data was X4, and the CPU usage data was Y4. After regression analysis and substitution into the comprehensive variable equation, the regression coefficient was 2.47E-05. For the mobile hall channel business volume data, the independent variable data was X5, and the CPU usage data was Y5. After regression analysis and substitution into the comprehensive variable equation, the regression coefficient was 1.88E-05. For other channel business volume data, the independent variable data was X6, and the CPU usage data was Y6. After regression analysis and substitution into the comprehensive variable equation, the regression coefficient was 1.90E-05. The t-value obtained after regression analysis and substitution into the comprehensive variable equation for the front-end channel business volume data, online hall channel business volume data, customer service channel business volume data, short-term hall channel business volume data, mobile hall channel business volume data, and other channel business volume data was 83.45, with a significance level of 0.000. The model summary R-value was 0.92, and the standard error was 7.7. The principal component regression model can be seen in Table 4. Table 4 is the principal component regression model.

[0084] Table 4 Principal Component Regression Model

[0085]

[0086] Finally, the management system constructs a performance capacity model based on the principal component regression model construction method. Specifically, the management system determines the hourly baseline peak business volume and individual CPU requirement (number of logical operations) corresponding to each business volume data in the principal component regression model according to the principal component regression model construction method, and constructs the corresponding performance capacity model based on the hourly baseline peak business volume and individual CPU requirement, combined with the principal component regression model.

[0087] In the above embodiments, based on the construction method of the principal component regression model, the hourly benchmark peak business volume data of the front-end channel is determined to be 218875, with a single CPU requirement of 4; the hourly benchmark peak business volume data of the online hall is 1305460, with a single CPU requirement of 33; the hourly benchmark peak business volume data of the customer service channel is 1513220, with a single CPU requirement of 41; the hourly benchmark peak business volume data of the short-term hall is 489965, with a single CPU requirement of 12; the hourly benchmark peak business volume data of the mobile hall is 360086, with a single CPU requirement of 7; and the hourly benchmark peak business volume data of other channels is 230998, with a single CPU requirement of 4. The performance capacity model is shown in Table 5 below. Table 5 is the performance capacity model.

[0088] Table 5 Performance Capacity Model

[0089]

[0090] In this embodiment, during the construction of the performance capacity model, the comprehensive variable equation is continuously optimized and trained through regression analysis, making the constructed performance capacity model more scientific and more accurate, and further improving the accuracy of performance capacity estimation.

[0091] This embodiment provides a method for constructing a performance capacity model. During the model construction process, no manual intervention is required. Principal component regression analysis is automatically used to analyze sample data and construct an accurate performance capacity model. This accurate model can accurately estimate the performance capacity of each product, improving the accuracy of performance capacity estimation. Furthermore, during the model construction process, principal component regression analysis is used to determine a comprehensive variable equation. This equation comprehensively represents the relationship between business volume data and CPU usage data, making the constructed performance capacity model more scientific and further improving the accuracy of performance capacity estimation.

[0092] Furthermore, referring to Figure 2 This is the second flowchart of the performance capacity model construction method provided in this application. After step S30, the method further includes:

[0093] Step S40: Based on the performance capacity model and according to the data relationship between the service ingress traffic data and the cluster traffic data of each cluster, estimate the estimated traffic data of each cluster.

[0094] Step S50: Estimate the module traffic data of each module in each cluster based on the estimated traffic data of each cluster and the number of machines corresponding to each cluster.

[0095] Step S60: Estimate the CPU usage data of each module in each cluster based on the preset module calculation formula and the module traffic data of each module in each cluster.

[0096] Step S70: Estimate the overall capacity based on the module CPU usage data of each module in each cluster.

[0097] It should be noted that the product is composed of the following components from top to bottom: product, service, cluster, and machine. Therefore, the total capacity of the product must ultimately be quantified onto the machine level. For ease of explanation, this embodiment uses a server as an example. A server is merely one manifestation of the product and does not limit the product itself. Therefore, it can also be understood that to determine the total capacity of the server, it is necessary to determine the capacity of each service within the server, and ultimately quantify the capacity of each service onto its corresponding machine level.

[0098] After training and obtaining the performance capacity model, the management system acquires the machine list of each cluster in the backend of each service. Based on the number of machines recorded in the machine list, it determines the number of machines in each cluster. It also acquires the module list from the machine list of each cluster and determines the number of modules in each cluster based on the number of modules recorded in the machine list. In this embodiment, if a cluster has 10 machines, and each machine corresponds to 10 modules, then the number of modules in that cluster is 10 * 10 = 100.

[0099] It should be noted that a server may contain multiple services. This embodiment uses a single service as an example and defines it as the target service. The management system uses a performance capacity model and estimates the overall machine capacity based on the target service's ingress traffic data and the number of modules in each cluster. Specifically, the management system calculates the cluster traffic data of each backend cluster of the target service, and then estimates the estimated traffic data of each cluster based on the ratio of the cluster traffic data to the ingress traffic data. In this embodiment, if the cluster traffic data of a certain cluster is 500, and the ratio of the traffic of this cluster to the ingress traffic data is 20%, then the estimated traffic data of this cluster is 500 * 20% = 100.

[0100] Next, the management system estimates the module traffic data of each module in each cluster based on the estimated traffic data of each cluster and the corresponding number of modules. Further, the management system substitutes the module traffic data of each module into a preset module calculation formula to estimate the module CPU usage data of each module in each cluster. This preset module calculation formula is set based on actual conditions. Finally, the management system estimates the capacity of the target service based on the module CPU usage data of each module in each cluster, and performs the operation on all services on the server in relation to the target service, determining the capacity of all services as the total capacity of the server, as described in steps S701 to S703.

[0101] Furthermore, the specific descriptions of steps S701 to S703 are as follows:

[0102] Step S701: Based on the process CPU usage data of the target machine and the sum of the module CPU usage data of all modules in the target machine, the first single-machine CPU usage data is obtained.

[0103] Step S702: Based on the process CPU usage data and the module CPU usage data of any module of the target machine, obtain the second single-machine CPU usage data;

[0104] Step S703: Estimate the total capacity of the machine based on the first single-machine CPU usage data and the second single-machine CPU usage data.

[0105] Specifically, the management system selects any machine in any cluster within the target service as the target machine and determines the CPU usage data of the processes surrounding the target machine. Next, the management system adds the CPU usage data of the processes surrounding the target machine to the sum of the CPU usage data of all modules within the target machine, obtaining a first calculation result, which is then determined as the first single-machine CPU usage data for the target machine. Simultaneously, the management system adds the CPU usage data of the processes surrounding the target machine to the sum of the CPU usage data of any module within the target machine, obtaining a second calculation result, which is then determined as the second single-machine CPU usage data for the target machine. Finally, the management system performs a comprehensive estimation based on the first and second single-machine CPU usage data to obtain the final single-machine CPU usage data for the target machine. Then, based on the final single-machine CPU usage data and the number of machines in the target service, the management system estimates the capacity of the target service and performs the same operations on all services within the server as described above, determining the capacity of all services as the estimated total machine capacity.

[0106] This application embodiment comprehensively estimates the overall CPU usage data of a single machine by combining the CPU usage data of all modules of a single machine and the CPU usage data of a single module, ensuring the accuracy of the CPU usage data of a single machine. Furthermore, it estimates the overall machine capacity based on the final CPU usage data of a single machine, ensuring the accuracy of the overall machine capacity budget.

[0107] Furthermore, after obtaining the total capacity of the product, i.e. the product's traffic, the management system estimates the product's capacity utilization rate based on the proportion of the product's traffic in the system's burn-down chart, and further improves the product based on the capacity utilization rate.

[0108] Furthermore, after obtaining the traffic data for each product, the management system acquires the traffic data for all current services of each product and determines whether a corresponding number of products need to be delisted based on the current traffic data for all current services. In this embodiment, for example, if the total capacity of both product A and product B is 100, the current traffic data for all current services of product A is 30, and the current traffic data for all current services of product B is 20, then all current services in product A can be migrated to product B, or all current services in product B can be migrated to product A, saving resource utilization and cost budget.

[0109] This embodiment provides a method for constructing a performance capacity model. By constructing the performance capacity model, the total capacity of a product is estimated, accurately predicting the total capacity of each product. Based on the accurate total capacity, every portion of the capacity resources in each product is fully utilized, better controlling costs while meeting business needs. Simultaneously, operations and maintenance personnel can clearly understand the maximum pressure and critical values ​​of the system's traffic capacity, and clearly understand whether the system needs to add new products, thereby ensuring that the system remains operational and improving system performance.

[0110] Furthermore, after step S70, the following steps are also included:

[0111] Step S80: Determine the resource usage data of the newly added service volume data based on the access traffic data of the newly added service volume data and the overall machine capacity.

[0112] Specifically, when the management system detects new service traffic data access, it determines the access traffic data of the new service traffic data and, based on this access traffic data, determines the required capacity. Next, the management system determines whether there is remaining capacity in the current product's overall capacity. If there is remaining capacity, it determines whether the remaining capacity is greater than the required capacity. If the remaining capacity is greater than the required capacity, the corresponding required capacity will be allocated from the remaining capacity for the new service traffic data. If the remaining capacity is less than the required capacity, it determines whether the required capacity is greater than the overall capacity. If the required capacity is less than the overall capacity, a new product is added, and the corresponding required capacity is allocated to the new product. If the required capacity is greater than the overall capacity, a corresponding number of products are added based on the multiple relationship between the required capacity and the overall capacity, and the corresponding required capacity is allocated to the new service traffic data in the multiple new products. In this embodiment, for example, if the total capacity of product A is 100, and the capacity required for the new business volume data is 250, then 3 new products A need to be added for the new business volume data.

[0113] This embodiment can accurately determine whether a corresponding number of products need to be added based on the access traffic data of new service volume and the overall capacity of the products. This ensures that the system will not crash due to capacity constraints, thus guaranteeing that the system can always operate and improving system performance. Simultaneously, when a new project is launched, the access traffic data of the new project and the overall capacity of the products are used to determine whether the new project will put pressure on the online services of the products, thereby ensuring the normal operation of online services and improving system performance.

[0114] Furthermore, the performance capacity model construction apparatus provided in this application will be described below. The performance capacity model construction apparatus described below can be referred to in correspondence with the performance capacity model construction method described above.

[0115] like Figure 3 As shown, Figure 3 This is a schematic diagram of the performance capacity model building device provided in this application. The performance capacity model building device includes:

[0116] The synthesis module 301 is used to synthesize corresponding sample data based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data;

[0117] Analysis module 302 is used to analyze the sample data according to a preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0118] The construction module 303 is used to train the comprehensive variable equation with the comprehensive variable as the independent variable data and the second CPU usage data to construct the corresponding performance capacity model.

[0119] Furthermore, the analysis module 302 is also used for:

[0120] After removing the CPU usage data consumed by the background monitoring business, fault handling business and upgrade and maintenance business in the sample data, the principal component data corresponding to the sample data is obtained;

[0121] Dimensionality reduction analysis is performed on the principal component data based on a preset analysis matrix to obtain the target principal component data and its corresponding component matrix.

[0122] Based on the target principal component data and the component matrix, the comprehensive variable equation and its corresponding comprehensive variables are determined.

[0123] Furthermore, the analysis module 302 is also used for:

[0124] The business volume data in the component matrix is ​​multiplied one-to-one with the CPU usage data corresponding to the target principal component data in the component matrix to construct the corresponding calculation equation;

[0125] The calculation equation is determined as the comprehensive variable equation, and the equation result corresponding to the comprehensive variable equation is determined as the comprehensive variable.

[0126] Furthermore, the construction module 303 is also used for:

[0127] Regression analysis was performed using the composite variable as the independent variable and the second CPU data to obtain the corresponding regression results;

[0128] The regression results are used to train the comprehensive variable equation to construct the corresponding principal component regression model, and the performance capacity model is constructed based on the principal component regression model.

[0129] Furthermore, the performance capacity model construction apparatus further includes: an estimation module, the estimation module being used for:

[0130] Based on the performance capacity model and the data relationship between the service ingress traffic data and the cluster traffic data of each cluster, the estimated traffic data of each cluster is estimated.

[0131] Based on the estimated traffic data of each cluster and the corresponding number of machines, estimate the module traffic data of each module in each cluster.

[0132] Based on the preset module calculation formula and the module traffic data of each module in each cluster, estimate the module CPU usage data of each module in each cluster.

[0133] The overall system capacity is estimated based on the CPU usage data of each module in each cluster.

[0134] Furthermore, the estimation module is also used for:

[0135] The first single-machine CPU usage data is obtained based on the process CPU usage data of the target machine and the sum of the module CPU usage data of all modules in the target machine.

[0136] Based on the process CPU usage data and the module CPU usage data of any module of the target machine, the second single-machine CPU usage data is obtained.

[0137] The overall capacity of the machine is estimated based on the first single-machine CPU usage data and the second single-machine CPU usage data.

[0138] Furthermore, the estimation module is also used for:

[0139] Based on the access traffic data of the newly added service volume data and the overall machine capacity, the resource usage data of the newly added service volume data is determined.

[0140] The specific embodiments of the performance capacity model construction apparatus provided in this application are basically the same as the embodiments of the performance capacity model construction method described above, and will not be repeated here.

[0141] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a performance capacity model construction method, which includes:

[0142] Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, corresponding sample data are synthesized.

[0143] The sample data are analyzed according to the preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0144] The comprehensive variable equation is trained using the comprehensive variable as the independent variable data and the second CPU usage data to construct the corresponding performance capacity model.

[0145] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0146] On the other hand, this application also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to execute the performance capacity model construction method provided by the above methods, the method including:

[0147] Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, corresponding sample data are synthesized.

[0148] The sample data are analyzed according to the preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0149] The comprehensive variable equation is trained using the comprehensive variable as the independent variable data and the second CPU usage data to construct the corresponding performance capacity model.

[0150] Furthermore, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the performance capacity model construction methods provided above, the method comprising:

[0151] Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, corresponding sample data are synthesized.

[0152] The sample data are analyzed according to the preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables;

[0153] The comprehensive variable equation is trained using the comprehensive variable as the independent variable data and the second CPU usage data to construct the corresponding performance capacity model.

[0154] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0155] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for constructing a performance capacity model, characterized in that, include: Based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data, corresponding sample data are synthesized. The sample data is analyzed using a preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables. This includes: removing CPU usage data consumed by background monitoring, fault handling, and upgrade maintenance services from the sample data to obtain the principal component data corresponding to the sample data; performing dimensionality reduction analysis on the principal component data based on a preset analysis matrix to obtain the target principal component data and its corresponding component matrix; and determining the comprehensive variable equation and its corresponding comprehensive variables based on the target principal component data and the component matrix. The comprehensive variable equation is trained using comprehensive variable data as independent variables and second CPU usage data to construct the corresponding performance capacity model; The step of determining the comprehensive variable equation and its corresponding comprehensive variable based on the target principal component data and the component matrix includes: multiplying the business volume data in the component matrix with the CPU usage data corresponding to the target principal component data in the component matrix one-to-one to construct the corresponding calculation equation; determining the calculation equation as the comprehensive variable equation, and determining the equation result corresponding to the comprehensive variable equation as the comprehensive variable.

2. The performance capacity model construction method according to claim 1, characterized in that, The steps of training the comprehensive variable equation using comprehensive variable data and second CPU usage data to construct the corresponding performance capacity model include: Regression analysis was performed using the composite variable as the independent variable and the second CPU data to obtain the corresponding regression results; The regression results are used to train the comprehensive variable equation to construct the corresponding principal component regression model, and the performance capacity model is constructed based on the principal component regression model.

3. The performance capacity model construction method according to claim 1, characterized in that, After the step of training the comprehensive variable equation with comprehensive variable data and second CPU usage data to construct the corresponding performance capacity model, the method further includes: Based on the performance capacity model and the data relationship between the service ingress traffic data and the cluster traffic data of each cluster, the estimated traffic data of each cluster is estimated. Based on the estimated traffic data of each cluster and the corresponding number of machines, estimate the module traffic data of each module in each cluster. Based on the preset module calculation formula and the module traffic data of each module in each cluster, estimate the module CPU usage data of each module in each cluster. The overall system capacity is estimated based on the CPU usage data of each module in each cluster.

4. The performance capacity model construction method according to claim 3, characterized in that, The step of estimating the overall system capacity based on the module CPU usage data of each module in each cluster includes: The first single-machine CPU usage data is obtained based on the process CPU usage data of the target machine and the sum of the module CPU usage data of all modules in the target machine. Based on the process CPU usage data and the module CPU usage data of any module of the target machine, the second single-machine CPU usage data is obtained. The overall system capacity is estimated based on the first single-machine CPU usage data and the second single-machine CPU usage data.

5. The performance capacity model construction method according to any one of claims 3 to 4, characterized in that, After the step of estimating the overall machine capacity based on the module CPU usage data of each module in each of the clusters, the method further includes: Based on the access traffic data of the newly added service volume data and the overall machine capacity, the resource usage data of the newly added service volume data is determined.

6. A performance capacity model construction device, characterized in that, include: The synthesis module is used to synthesize corresponding sample data based on the correlation between the business volume data of each interface channel and its corresponding first CPU usage data; The analysis module is used to analyze the sample data according to a preset principal component regression analysis method to obtain the corresponding comprehensive variable equation and its corresponding comprehensive variables. This includes: removing CPU usage data consumed by background monitoring, fault handling, and upgrade / maintenance services from the sample data to obtain the principal component data corresponding to the sample data; performing dimensionality reduction analysis on the principal component data based on a preset analysis matrix to obtain target principal component data and its corresponding component matrix; and determining the comprehensive variable equation and its corresponding comprehensive variables based on the target principal component data and the component matrix. The module is used to train the equation of the comprehensive variable with the comprehensive variable as the independent variable data and the data used by the second CPU, and to build the corresponding performance capacity model. The step of determining the comprehensive variable equation and its corresponding comprehensive variable based on the target principal component data and the component matrix includes: multiplying the business volume data in the component matrix with the CPU usage data corresponding to the target principal component data in the component matrix one-to-one to construct the corresponding calculation equation; determining the calculation equation as the comprehensive variable equation, and determining the equation result corresponding to the comprehensive variable equation as the comprehensive variable.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the performance capacity model construction method according to any one of claims 1 to 5.

8. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the performance capacity model construction method according to any one of claims 1 to 5.

Images