Computer performance evaluation system and method based on big data and AI

Abstract

The invention relates to the technical field of performance evaluation, in particular to a computer performance evaluation system and method based on big data and AI. The system comprises a request processing unit, a calculation time unit, a subset influence unit and a performance evaluation unit, and the subset influence unit calculates response time length subset feature values according to extracted feature values and then calculates the comprehensive influence degree according to the calculated response time length subset feature values. The performance evaluation unit evaluates the computer performance delay probability according to the calculated comprehensive influence degree and the calculated average response time, and when the evaluated performance delay probability is larger than 1, it shows that the influence of the extracted feature value on the calculated average response time in the response time long subset is obvious; the computer performance delay is evaluated according to the comprehensive influence degree, the obvious influence of the features on the response time can be accurately recognized, the influence of subjective judgment is reduced, and more accurate delay probability evaluation is provided.

Description

Technical Field

[0001] The present invention relates to the field of performance evaluation technology, and in particular to a computer performance evaluation system and method based on big data and AI. Background Art

[0002] Computer performance evaluation based on big data and AI combines big data technology and artificial intelligence, and can conduct a comprehensive analysis of computer performance to help enterprises and organizations optimize system performance, prevent potential problems and improve business efficiency. When we interact with computer AI, computer AI will collect and process a large amount of request data. Since processing a large amount of request data will lead to slow request processing speed, it is impossible to accurately identify whether the requested data has a significant impact on the response time when the request processing speed is slow, which increases the complexity of the system. At the same time, it is difficult to determine the specific contribution of the data to the response time, which leads to a decrease in the accuracy of the delay probability of computer performance evaluation. Therefore, we provide a computer performance evaluation system and method based on big data and AI. Summary of the invention

[0003] The purpose of the present invention is to provide a computer performance evaluation system and method based on big data and AI to solve the problems raised in the above background technology.

[0004] To achieve the above-mentioned object, one of the objects of the present invention is to provide a computer performance evaluation system based on big data and AI, comprising a request processing unit, a computing time unit, a subset impact unit, and a performance evaluation unit; The request processing unit integrates the data in the log records into a data set, calculates the request processing speed, and determines that the request processing speed is slow using the calculated request processing speed; The calculation time unit is used to receive data from the request processing unit, randomly select data from the data set as the initial subset center and then calculate the distance, distribute the data to the subset corresponding to the initial subset center with the shortest distance according to the distance, record the data response time in the subset, calculate the new subset center, update the subset according to the calculated new subset center, record the new data response time in the updated subset, calculate the average response time according to the new data response time, and identify the subset with a long average response time; The subset impact unit is used to receive data from the request processing unit and the calculation time unit, extract feature values ​​from the data in the long response time subset, calculate feature values ​​and impact of the long response time subset according to the extracted feature values, and then calculate the comprehensive impact of the long response time subset according to the calculated feature values ​​of the long response time subset; The performance evaluation unit is used to receive data in the subset impact unit and the calculation time unit, and evaluate the computer performance delay probability according to the calculated comprehensive impact degree and the calculated average response time.

[0005] As a further improvement of the present technical solution, the request processing unit obtains the request time, request type, response time, and request source through log records, then processes the request data, and records the throughput and request processing time, calculates the request processing speed based on the throughput and request processing time, and uses the calculated request processing speed and the set request processing speed threshold to make a judgment to determine whether the calculated request processing speed is lower than the set request processing speed threshold. When the calculated request processing speed is lower than the set request processing speed threshold, it indicates that the request processing speed is slow.

[0006] As a further improvement of the technical solution, the computing time unit includes a computing allocation module, a central updating module and a response time module; The computing allocation module is used to receive a command indicating that the request processing speed is slow in the request processing unit. The computing allocation module obtains a data set from the request processing unit, divides the data set into three subsets, and randomly selects data from the data set as the initial subset center. Then, data is randomly obtained from the data set, and feature values ​​are extracted from the randomly obtained data and the initial subset center using big data technology, and the number of features is recorded to obtain the extracted data feature values ​​and the extracted initial subset center feature values. Then, the distance is calculated based on the extracted data feature values, the extracted initial subset center feature values, and the number of features, and the distance from the randomly obtained data to different initial subset centers is calculated. Then, the distance from the randomly obtained data to different initial subset centers is compared in turn, and the shortest distance is selected. According to the lowest distance, the randomly obtained data is allocated to the subset corresponding to the initial subset center with the shortest distance, and the number of data in the subset and the data response time are recorded. The center update module is used to receive the number of subsets and data in the subsets in the calculation allocation module, calculate the new subset center according to the number of subsets and data in the subsets, compare the calculated new subset center with the set subset center threshold, and when the calculated new subset center is less than the set subset center threshold, update the calculated new subset center to obtain an updated subset, and record the number of new data in the updated subset and the new data response time; The response time module is used to receive update commands in the central update module. The response time module obtains the number of new data and the new data response time in the updated subset from the central update module, calculates the average response time based on the number of new data and the new data response time, calculates the average response time of different subsets, and uses the calculated average response times of different subsets to compare them in turn to identify the subset with the long calculated average response time.

[0007] As a further improvement of the technical solution, the average response time is calculated in the response time module according to the following principles: The amount of new data collected and new data response time Calculate the average response time and get the calculated average response time , the specific algorithm formula is: ; in, Refers to the The new data response time solves the problem of long average response time in the subset.

[0008] As a further improvement of the technical solution, the subset impact unit includes a subset feature module, an impact degree module and a comprehensive impact module; The subset feature module is used to receive the long response time subset in the response time module, extract feature values ​​from the data in the long response time subset using big data technology, and then record the number of data in the long response time subset, and calculate the feature value of the long response time subset based on the extracted feature values ​​and the number of data in the long response time subset; The influence module is used to receive the feature values ​​extracted from the subset feature module, the long response time subset and the data set in the request processing unit, use linear regression to establish a regression model based on the data, input the data in the long response time subset into the regression model, the regression model outputs the corresponding model coefficients, and then calculates the influence based on the model coefficients and the extracted feature values; The comprehensive impact module is used to receive the model coefficient in the impact module and the long response time subset characteristic value calculated in the subset characteristic module, and calculate the comprehensive impact of the long response time subset according to the model coefficient and the calculated long response time subset characteristic value.

[0009] As a further improvement of the technical solution, the algorithm formula for calculating the long subset characteristic value of the response time in the subset characteristic module is: Collect the extracted feature values and the number of data in the subset with long response time Calculate the long subset eigenvalue of the response time and obtain the calculated long subset eigenvalue of the response time , specific algorithm: ; in, Refers to the first The extracted feature values.

[0010] As a further improvement of the technical solution, the implementation principle of calculating the comprehensive influence degree in the comprehensive influence module is: Collect model coefficients And the calculated response time long subset eigenvalue Calculate the comprehensive impact and obtain the calculated comprehensive impact , the specific algorithm formula is: ; in, refers to the number of eigenvalues ​​of the long subset of response times calculated, Refers to the model coefficients.

[0011] As a further improvement of the present technical solution, the performance evaluation unit is used to receive the comprehensive impact calculated in the comprehensive impact module and the average response time calculated in the response time module, receive the characteristic values ​​extracted in the impact module and the long response time subset, and evaluate the probability of computer performance delay based on the calculated comprehensive impact and the calculated average response time. When the evaluated performance delay probability is greater than 1, it means that the extracted characteristic values ​​have a significant impact on the calculated average response time in the long response time subset, and there is a delay effect on the computer performance.

[0012] A second object of the present invention is to provide a method for operating a computer performance evaluation system based on big data and AI as described in any one of the above, comprising the following method steps: S1. The request processing unit integrates the data in the log records into a data set, calculates the request processing speed, determines that the request processing speed is slow by using the calculated request processing speed, and transmits the command of the slow request processing speed to the calculation time unit. The calculation time unit randomly selects data from the data set as the initial subset center and then calculates the distance. According to the distance, the data is distributed to the subset corresponding to the initial subset center with the shortest distance, and a new subset center is calculated. The subset is updated according to the calculated new subset center, and the response time of the new data in the updated subset is recorded at the same time. The average response time is calculated according to the new data response time, and the subset with a long average response time is identified; S2, the subset impact unit is used to receive the data in the request processing unit and the calculation time unit, extract the characteristic value of the data in the long response time subset, calculate the characteristic value and influence of the long response time subset according to the extracted characteristic value, and then calculate the comprehensive influence according to the calculated characteristic value of the long response time subset; S3. The performance evaluation unit is used to receive data from the subset impact unit and the calculation time unit, and evaluate the probability of computer performance delay according to the calculated comprehensive impact and the calculated average response time.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. In the computer performance evaluation system and method based on big data and AI, the performance evaluation unit evaluates the probability of computer performance delay according to the calculated comprehensive impact and the calculated average response time. When the evaluated performance delay probability is greater than 1, it means that the extracted feature value has a significant impact on the calculated average response time in the long response time subset, and the computer performance has a delay effect. By evaluating the computer performance delay according to the comprehensive impact, it is possible to accurately identify that the feature has a significant impact on the response time, reduce the impact of subjective judgment, and provide a more accurate delay probability evaluation.

[0014] 2. In the computer performance evaluation system and method based on big data and AI, the response time module obtains the number of new data and the response time of the new data in the updated subset from the central update module, calculates the average response time according to the number of new data and the response time of the new data, calculates the average response time of different subsets, and compares the average response time of different subsets in turn to identify the subset with a long average response time. Identifying the subset with a long average response time can accurately locate the performance bottleneck area in the computer, and then concentrate resources on the bottleneck area, including increasing computing power, optimizing database queries, improving network performance, and solving problems that lead to performance degradation. It is no longer necessary to optimize all data equally, thereby avoiding waste of resources in the normal performance part and improving resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is an overall block diagram of the present invention; Figure 2 It is a block diagram of the computing time unit of the present invention; Figure 3 A block diagram of a subset of the present invention affecting the unit; Figure 4 It is a flowchart of the steps of the present invention.

[0016] The meaning of each number in the figure is: 1. Request processing unit; 2. Calculation time unit; 21. Calculation allocation module; 22. Center update module; 23. Response time module; 3. Subset impact unit; 31. Subset feature module; 32. Impact module; 33. Comprehensive impact module; 4. Performance evaluation unit. DETAILED DESCRIPTION

[0017] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0018] See also Figure 1-Figure 3 As shown, one of the purposes of this embodiment is to provide a computer performance evaluation system based on big data and AI, including a request processing unit 1, a computing time unit 2, a subset impact unit 3, and a performance evaluation unit 4; The request processing unit 1 integrates the data in the log records into a data set, calculates the request processing speed, and determines that the request processing speed is slow by using the calculated request processing speed; The calculation time unit 2 is used to receive the data in the request processing unit 1, randomly select data from the data set as the initial subset center and then calculate the distance, distribute the data to the subset corresponding to the initial subset center with the shortest distance according to the distance, and record the data response time in the subset, calculate the new subset center, update the subset according to the calculated new subset center, and record the new data response time in the updated subset, calculate the average response time according to the new data response time, and identify the subset with a long average response time; The subset impact unit 3 is used to receive data from the request processing unit 1 and the calculation time unit 2, extract feature values ​​from the data in the long response time subset, calculate feature values ​​and impact of the long response time subset according to the extracted feature values, and then calculate the comprehensive impact of the long response time subset according to the calculated feature values ​​of the long response time subset; The performance evaluation unit 4 is used to receive data from the subset impact unit 3 and the calculation time unit 2, and evaluate the computer performance delay probability according to the calculated comprehensive impact degree and the calculated average response time.

[0019] The following is a refinement of the above units, see Figure 2-Figure 4 As shown; The request processing unit 1 obtains the request time, request type, response time, and request source through log records, integrates the obtained request time, request type, response time, and request source into a data set, processes the request data, and records the throughput (the number of requests processed per unit time) and the request processing time, calculates the request processing speed according to the throughput and the request processing time, and uses the calculated request processing speed and the set request processing speed threshold to judge whether the calculated request processing speed is lower than the set request processing speed threshold. When the calculated request processing speed is lower than the set request processing speed threshold, it indicates that the request processing speed is slow, and the command indicating the slow request processing speed is transmitted to the calculation time unit 2; The implementation principle of calculating request processing speed: Get Throughput and request processing time Calculate the request processing speed and get the calculated request processing speed , the specific algorithm formula is: ; The calculation time unit 2 includes a calculation allocation module 21, a central update module 22 and a response time module 23; The calculation allocation module 21 is used to receive the command of slow request processing speed in the request processing unit 1. The calculation allocation module 21 obtains a data set from the request processing unit 1, divides the data set into 3 subsets, and randomly selects data from the data set as the initial subset center. Then, randomly obtains data from the data set, and uses big data technology to extract feature values ​​of the randomly obtained data and the initial subset center, and records the number of features to obtain the extracted data feature values ​​and the extracted initial subset center feature values. Then, the distance is calculated based on the extracted data feature values, the extracted initial subset center feature values ​​and the number of features, and so on. The distance from the randomly obtained data to different initial subset centers is calculated, and then the distances from the randomly obtained data to different initial subset centers are compared in turn, the shortest distance is selected, and the randomly obtained data is allocated to the subset corresponding to the initial subset center with the shortest distance according to the lowest distance, and the number of data in the subset and the data response time are recorded at the same time; The implementation principle of calculating distance: Collect the extracted data feature values , the extracted initial subset center eigenvalue and the number of features Calculate the distance and get the calculated distance , distance algorithm formula: ; in, Refers to the The extracted data feature values, Refers to the The extracted initial subset center eigenvalue; The center update module 22 is used to receive the number of subsets and data in the subsets in the calculation allocation module 21, calculate the new subset center according to the number of subsets and data in the subsets, compare the calculated new subset center with the set subset center threshold, and when the calculated new subset center is less than the set subset center threshold, update the calculated new subset center to obtain an updated subset, and record the number of new data in the updated subset and the new data response time; Principle of calculating the center of the new subset: Collecting Subsets and the number of data in the subset Calculate the new subset center and get the calculated new subset center , the specific algorithm formula is: ; in, refers to the subset individual data; The response time module 23 is used to receive the updated command in the central update module 22. The response time module 23 obtains the number of new data and the new data response time in the updated subset from the central update module 22, calculates the average response time according to the number of new data and the new data response time, and so on, calculates the average response time of different subsets, and compares the average response time of different subsets in turn to identify the subset with a long average response time. Identifying the subset with a long average response time can help accurately locate the performance bottleneck area in the computer, which helps to focus on solving the problems that cause performance degradation instead of processing all data, and the subset with a long response time is passed to the subset impact unit 3; The implementation principle of calculating the average response time in the response time module 23 is: The amount of new data collected and new data response time Calculate the average response time and get the calculated average response time , the specific algorithm formula is: ; in, Refers to the The new data response time solves the problem of long average response time in the subset, which can reduce the fluctuation of computer performance and improve overall stability; The subset impact unit 3 includes a subset feature module 31, an impact degree module 32 and a comprehensive impact module 33; The subset feature module 31 is used to receive the long response time subset in the response time module 23, extract feature values ​​from the data in the long response time subset using big data technology, and then record the number of data in the long response time subset, calculate feature values ​​of the long response time subset based on the extracted feature values ​​and the number of data in the long response time subset, and use the data in the long response time subset to calculate feature values, so as to more accurately select features that need to be optimized; The algorithm formula for calculating the long subset characteristic value of the response time in the subset characteristic module 31 is: Collect the extracted feature values and the number of data in the subset with long response time Calculate the long subset eigenvalue of the response time and obtain the calculated long subset eigenvalue of the response time , specific algorithm: ; in, Refers to the first The extracted feature values; The impact module 32 is used to receive the feature values ​​extracted in the subset feature module 31, the long response time subset and the data set in the request processing unit 1, and use linear regression to establish a regression model based on the data, input the data in the long response time subset into the regression model, and the regression model outputs the corresponding model coefficients, and then calculates the impact based on the model coefficients and the extracted feature values. Identifying high-impact features in advance can help take measures before problems occur and reduce the risk of computer performance fluctuations; Principle of calculating influence: Collect model coefficients and the extracted eigenvalues Calculate the influence and get the calculated influence , the specific algorithm formula is: ; in, Refers to the The model coefficients, Refers to the first The extracted feature values; The comprehensive impact module 33 is used to receive the model coefficient in the impact module 32 and the long response time subset characteristic value calculated in the subset characteristic module 31, and calculate the comprehensive impact of the long response time subset according to the model coefficient and the calculated long response time subset characteristic value. By understanding the characteristics that lead to the long response time, the fluctuation of computer performance can be reduced and the overall stability can be improved; The implementation principle of calculating the comprehensive impact degree in the comprehensive impact module 33 is: Collect model coefficients And the calculated response time long subset eigenvalue Calculate the comprehensive impact and obtain the calculated comprehensive impact , the specific algorithm formula is: ; in, refers to the number of eigenvalues ​​of the long subset of response times calculated, Refers to the Through the calculation of the comprehensive influence of the model coefficients, it can be clearly seen that the extracted features have a significant impact on the system performance in the long response time subset, making it more accurate to locate and solve performance bottlenecks; The performance evaluation unit 4 is used to receive the comprehensive impact calculated in the comprehensive impact module 33 and the average response time calculated in the response time module 23, receive the feature value extracted in the impact module 32, and the long response time subset, and evaluate the probability of computer performance delay according to the calculated comprehensive impact and the calculated average response time. When the evaluated performance delay probability is greater than 1, it means that the extracted feature value has a significant impact on the calculated average response time in the long response time subset, and the computer performance has a delay effect. By evaluating the computer performance delay according to the comprehensive impact, it is possible to accurately identify that the feature has a significant impact on the response time, reduce the impact of subjective judgment, and provide a more accurate delay probability evaluation; The implementation principle of evaluating the probability of computer performance delay: Collect and calculate the comprehensive impact and calculate the average response time Calculate the probability of computer performance delay. The calculated probability of computer performance delay is the evaluated probability of computer performance delay. , the specific algorithm formula is: .

[0020] The second object of the present invention is to provide a method for operating the above-mentioned computer performance evaluation system based on big data and AI, please refer to Figure 4 As shown, the method comprises the following steps: S1, request processing unit 1 integrates the data in the log records into a data set, calculates the request processing speed, uses the calculated request processing speed to determine that the request processing speed is slow, and transmits the command of the slow request processing speed to the calculation time unit 2. The calculation time unit 2 randomly selects data from the data set as the initial subset center and then calculates the distance. According to the distance, the data is distributed to the subset corresponding to the initial subset center with the shortest distance, and the new subset center is calculated. The subset is updated according to the calculated new subset center, and the response time of the new data in the updated subset is recorded at the same time. The average response time is calculated according to the new data response time, and the subset with a long average response time is identified; S2, the subset impact unit 3 is used to receive the data in the request processing unit 1 and the calculation time unit 2, extract the characteristic value of the data in the long response time subset, calculate the characteristic value and influence of the long response time subset according to the extracted characteristic value, and then calculate the comprehensive influence according to the calculated characteristic value of the long response time subset; S3. The performance evaluation unit 4 is used to receive the data in the subset impact unit 3 and the calculation time unit 2, and evaluate the computer performance delay probability according to the calculated comprehensive impact degree and the calculated average response time.

[0021] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A computer performance evaluation system based on big data and AI, characterized by: It includes a request processing unit (1), a calculation time unit (2), a subset impact unit (3), and a performance evaluation unit (4); The request processing unit (1) integrates the data in the log records into a data set, calculates the request processing speed, and determines that the request processing speed is slow based on the calculated request processing speed; The time calculation unit (2) is used to receive data from the request processing unit (1), randomly select data from the data set as the initial subset center and then calculate the distance, distribute the data to the subset corresponding to the initial subset center with the shortest distance according to the distance, record the data response time in the subset, calculate the new subset center, update the subset according to the calculated new subset center, record the new data response time in the updated subset, calculate the average response time according to the new data response time, and identify the subset with the long average response time; The subset impact unit (3) is used to receive data from the request processing unit (1) and the calculation time unit (2), extract characteristic values ​​from the data in the long response time subset, calculate characteristic values ​​and impact of the long response time subset based on the extracted characteristic values, and then calculate the comprehensive impact of the long response time subset based on the calculated characteristic values ​​of the long response time subset; The performance evaluation unit (4) is used to receive data from the subset impact unit (3) and the calculation time unit (2), and evaluate the computer performance delay probability according to the calculated comprehensive impact degree and the calculated average response time.

2. The computer performance evaluation system based on big data and AI according to claim 1, characterized in that: The request processing unit (1) obtains the request time, request type, response time, and request source through log records, processes the request data, and records the throughput and request processing time, calculates the request processing speed based on the throughput and request processing time, and uses the calculated request processing speed and a set request processing speed threshold to make a judgment to determine whether the calculated request processing speed is lower than the set request processing speed threshold. When the calculated request processing speed is lower than the set request processing speed threshold, it indicates that the request processing speed is slow.

3. The computer performance evaluation system based on big data and AI according to claim 2, characterized in that: The computing time unit (2) comprises a computing allocation module (21), a central updating module (22) and a response time module (23); The computing allocation module (21) is used to receive a command indicating a slow request processing speed in the request processing unit (1). The computing allocation module (21) obtains a data set from the request processing unit (1), divides the data set into three subsets, and randomly selects data from the data set as an initial subset center. Then, data is randomly obtained from the data set. The feature values ​​of the randomly obtained data and the initial subset center are extracted using big data technology, and the feature quantity is recorded to obtain the extracted data feature value and the extracted initial subset center feature value. Then, the distance is calculated based on the extracted data feature value, the extracted initial subset center feature value and the feature quantity, and the distance from the randomly obtained data to different initial subset centers is calculated. Then, the distance from the randomly obtained data to different initial subset centers is compared in turn, and the shortest distance is selected. According to the shortest distance, the randomly obtained data is allocated to the subset corresponding to the initial subset center with the shortest distance, and the number of data in the subset and the data response time are recorded. The center update module (22) is used to receive the number of subsets and data in the subsets in the calculation allocation module (21), calculate a new subset center according to the number of subsets and data in the subsets, compare the calculated new subset center with a set subset center threshold, and when the calculated new subset center is less than the set subset center threshold, update the calculated new subset center to obtain an updated subset, and record the number of new data in the updated subset and the new data response time; The response time module (23) is used to receive an update command from the central update module (22). The response time module (23) obtains the number of new data in the updated subset and the new data response time from the central update module (22), calculates the average response time based on the number of new data and the new data response time, calculates the average response time of different subsets, and compares the calculated average response times of different subsets in turn to identify a subset with a longer calculated average response time.

4. The computer performance evaluation system based on big data and AI according to claim 3, characterized in that: The principle of calculating the average response time in the response time module (23) is as follows: The amount of new data collected and new data response time Calculate the average response time and get the calculated average response time , the specific algorithm formula is: ； in, Refers to the The new data response time solves the problem of long average response time in the subset.

5. The computer performance evaluation system based on big data and AI according to claim 4, characterized in that: The subset impact unit (3) comprises a subset feature module (31), an impact degree module (32) and a comprehensive impact module (33); The subset feature module (31) is used to receive the long response time subset in the response time module (23), extract feature values ​​from the data in the long response time subset using big data technology, then record the number of data in the long response time subset, and calculate the feature value of the long response time subset based on the extracted feature values ​​and the number of data in the long response time subset; The influence degree module (32) is used to receive the feature values ​​extracted in the subset feature module (31), the long response time subset and the data set in the request processing unit (1), establish a regression model based on the data using linear regression, input the data in the long response time subset into the regression model, the regression model outputs the corresponding model coefficients, and then calculates the influence degree based on the model coefficients and the extracted feature values; The comprehensive impact module (33) is used to receive the model coefficient in the impact module (32) and the long response time subset characteristic value calculated in the subset characteristic module (31), and calculate the comprehensive impact of the long response time subset based on the model coefficient and the calculated long response time subset characteristic value.

6. The computer performance evaluation system based on big data and AI according to claim 5, characterized in that: The algorithm formula for calculating the long subset characteristic value of the response time in the subset characteristic module (31) is: Collect the extracted feature values and the number of data in the subset with long response time Calculate the long subset eigenvalue of the response time and obtain the calculated long subset eigenvalue of the response time , specific algorithm: ； in, Refers to the first The extracted feature values.

7. The computer performance evaluation system based on big data and AI according to claim 5, characterized in that: The implementation principle of calculating the comprehensive impact degree in the comprehensive impact module (33) is: Collect model coefficients And the calculated response time long subset eigenvalue Calculate the comprehensive impact and obtain the calculated comprehensive impact , the specific algorithm formula is: ； in, refers to the number of eigenvalues ​​of the long subset of response times calculated, Refers to the model coefficients.

8. The computer performance evaluation system based on big data and AI according to claim 4, characterized in that: The performance evaluation unit (4) is used to receive the comprehensive impact calculated in the comprehensive impact module (33) and the average response time calculated in the response time module (23), receive the characteristic value extracted in the impact module (32) and the long response time subset, and evaluate the probability of computer performance delay based on the calculated comprehensive impact and the calculated average response time. When the evaluated performance delay probability is greater than 1, it means that the extracted characteristic value has a significant impact on the calculated average response time in the long response time subset, and the computer performance has a delay effect.

9. A method for operating a computer performance evaluation system based on big data and AI according to any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1, the request processing unit (1) integrates the data in the log records into a data set, calculates the request processing speed, determines that the request processing speed is slow by using the calculated request processing speed, and transmits the command of the slow request processing speed to the calculation time unit (2). The calculation time unit (2) randomly selects data from the data set as the initial subset center and then calculates the distance. According to the distance, the data is allocated to the subset corresponding to the initial subset center with the shortest distance, and a new subset center is calculated. According to the calculated new subset center, the subset is updated, and the response time of the new data in the updated subset is recorded at the same time. According to the response time of the new data, the average response time is calculated, and the subset with the long average response time is identified; S2, the subset impact unit (3) is used to receive the data in the request processing unit (1) and the calculation time unit (2), extract the characteristic value of the data in the long response time subset, calculate the characteristic value and influence of the long response time subset according to the extracted characteristic value, and then calculate the comprehensive influence according to the calculated characteristic value of the long response time subset; S3, the performance evaluation unit (4) is used to receive the data in the subset impact unit (3) and the calculation time unit (2), and evaluate the computer performance delay probability according to the calculated comprehensive impact and the calculated average response time.

Images