CN110096335B - A prediction method of business concurrency for different types of virtual machines - Google Patents

Abstract

The invention provides a prediction method for different types of business concurrency of virtual machines, and relates to the technical field of cloud computing. A method for predicting business concurrency for different types of virtual machines. First, the historical business concurrency of virtual machines is collected and preprocessed, and then the virtual machine business concurrency is judged based on the improved 1-nearest neighbor-dynamic time adjustment method 1NN-DTW. Finally, a classification and regression tree is used to fit the business concurrency without periodic changes; Fourier series FS and a classification and regression tree CART are used to fit the business concurrency with periodic changes; It can provide a basis for the increase or decrease of virtual machines in the next step, and at the same time help to accurately estimate the software aging status of virtual machines, so as to improve the working virtual machine. performance and reliability purposes.

Description

A prediction method of business concurrency for different types of virtual machines

technical field

The invention relates to the technical field of cloud computing, and in particular, to a method for predicting business concurrency for different types of virtual machines.

Background technique

Software aging is common in cloud service systems. During the process of virtual machines processing concurrent business requests, operating systems and application software continue to accumulate errors, resulting in the gradual degradation of the performance of working virtual machines, which in turn affects the service quality of cloud service systems. . The highly scalable and dynamic reconfiguration characteristics of the cloud platform provide a technical basis for ensuring the quality of cloud services under different concurrency conditions. However, the existing dynamic adjustment methods for virtual resources still have many defects.

Generally speaking, a variety of services are deployed on virtual machines, and the change trend of the concurrency of each service at different times is different. For example, some services continue to increase during a certain period of time during the day, and continue to decrease during a certain period of time at night. , some business concurrency fluctuates continuously, while some business concurrency has remained stable. By predicting the concurrency of each business of the cloud platform, it can provide a basis for the increase or decrease of virtual machines in the next step, and at the same time help to accurately estimate the software aging status of virtual machines, so as to improve the performance and reliability of working virtual machines. .

Because user operations, virtual machine services, and other uncertain factors change all the time, the concurrent access volume of services not only changes steadily with time, but also tends to rise, fall, and cyclic fluctuations. Traditional load models such as exponential smoothing The model can only roughly describe the change trend of business concurrency, and cannot capture the nonlinear change characteristics well.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a method for predicting different types of business concurrency of virtual machines, so as to realize the prediction of different types of business concurrency in virtual machines.

A method for predicting business concurrency for different types of virtual machines, comprising the following steps:

Step 1: Collect historical business concurrency of virtual machines and preprocess them. The specific methods are as follows:

Step 1.1: Scan the business concurrency of virtual machines for a period of time to find the missing points of business concurrency;

Step 1.2: Process the scanned missing points of business concurrency;

Step 1.2.1: For the case of missing individual sampling points, use the average value of the business concurrency in the previous cycle and the next cycle to fill in. The calculation formula for the missing business concurrency con(t) in the t-th time period of the virtual machine is as follows: shown:

Step 1.2.2: For the case where the sample is missing more than 90%, discard all samples and set the value of business concurrency within this period of time to zero;

Step 1.3: Perform outlier adjustment for the extremely small samples with abnormal fluctuations in the collected business concurrency;

Step 1.3.1: Combine the quartiles to calculate the upper limit H and lower limit L of the normal value of the virtual machine business concurrency in time t, as shown in the following formula:

H=Q3+k*(Q3-Q1) (2)

L=Q1-k*(Q3-Q1) (3)

Among them, Q1 represents the lower quartile, that is, the 25th percentile of the ascending sequence of business concurrency in time t, and Q3 represents the upper quartile, that is, the percent of the ascending sequence of business concurrency in time t Seventy-five points, k is used to describe the abnormal degree of unreasonable sampling points, generally 1.5 and 3, representing moderate and extreme respectively;

Step 1.3.2: Determine whether the data of each sampling point is normal through the Tukey test method, and adjust the abnormal values;

If the data value of the sampling point is determined to be a sample of wrong business concurrency, the wrong value will be discarded first, and then supplemented by the mean value filling method;

If the data value of the sampling point is determined to be a sample of normal business concurrency, no adjustment will be made;

Step 1.4: Adjust the data interval for the business concurrency and CPU utilization data collected from the log database or management log, and merge the collected data in seconds, minutes or hours;

Step 1.5: Use the maximum and minimum normalization method to normalize the data processed in step 1.4;

Step 2: Determine the type of virtual machine business concurrency based on the improved 1-Nearest Neighbor-Dynamic TimeWarping (1-NearestNeighbor-Dynamic TimeWarping, ie 1NN-DTW) method. The specific method is as follows:

Step 2.1: Classify the concurrency of each service of the virtual machine into rising, falling, quadratic, random, cyclically fluctuating, cyclically rising and cyclically falling;

Step 2.2: For various types of business concurrency, select the labeled business concurrency sequence as a known sample in advance;

Step 2.3: For each sequence of business concurrency to be classified, scan all known samples in turn and calculate the closest known sample through the proximity algorithm, then the type of the known sample is the type of business concurrency to be classified ;

Step 2.4: Classify all business concurrency into two categories to simplify the 1-nearest neighbor model;

Classify random, rising, falling, and quadratic business concurrency into categories that do not have periodic changes;

Classify the cyclically fluctuating, cyclically rising, and cyclically falling business concurrency as categories with cyclical changes;

Step 2.5: Construct an n×m matrix so that the sequence of business concurrency to be classified {x ₁ , x ₂ , ..., x _n } and a known sequence of business concurrency {y ₁ , y ₂ , ... , y _m } are aligned, where n is the total number of business concurrency to be classified, and m is the known total number of business concurrency;

Step 2.6: Take the two-point deviation of the i-th business concurrency x _i to be classified and the known j-th business concurrency amount y _j as the value d _{i, j} of the (i, j) position in the matrix, and use the Euclidean distance at the same time And the method of the square of the difference of the two-point derivative to calculate the sequence of business concurrency to be classified {x ₁ , x ₂ , ..., x _n } and the known business concurrency sequence {y ₁ , y ₂ , ... , y _m } The deviation d _{i, j} of each point after alignment is shown in the following formula:

d _i,j =(x _i -y _j ) ² +(x' _i -y' _j ) ² (4)

Among them, x′ _i and y′ _j are the derivatives of x _i and y _j respectively, and the estimation of the derivative x′ _i of the service concurrency x _i is shown in the following formula:

Step 2.7: Starting from position (1, 1) in the matrix, iteratively find a path with the smallest cumulative deviation according to the constraints that each position can only reach the position above, right or upper right except for the boundary value, until the end of position (n, m);

Step 3: Predict the business concurrency of different types of changes in the virtual machine. The specific methods are:

Step 3.1: Use a Classification and Regression Tree (CART) to fit the business concurrency without periodic changes;

Step 3.1.1: Traverse the arbitrary value f of each feature F of the sample business concurrency sequence, divide the sample data with (F, f) as the condition, determine the division position with the smallest square error, and select the highest value from the business concurrency sequence. good cutting point;

The calculation formula of the squared error error is as follows:

代表样本x中第i’个业务并发量的特征，y_i，代表分割前的第i’个序列样本，

代表分割后的第i’个子序列样本的拟合结果；in,

Represents the feature of the i'th business concurrency in the sample x, y _i, represents the i'th sequence sample before splitting,

represents the fitting result of the i'th subsequence sample after segmentation;

Step 3.1.2: Save the business concurrency value as the cutting point, and perform segmentation on the business concurrency sequence;

Step 3.1.3: Construct subtrees with feature F greater than f and subtrees smaller than f in turn, and further iteratively segment and fit the sequence of business concurrency on the left and right of the current split point until it can no longer be divided into leaf nodes;

Step 3.1.4: Re-traverse the sample data from bottom to top, check each split point for all business concurrency series, and judge the fitting error of the concurrency series before and after split.

If the fitting error of the concurrency sequence decreases after the split, the split point is retained;

If the fitting error of the concurrent series increases after the split, cancel the split point and merge the left and right series;

Step 3.2: Use Fourier series FS and classification regression tree CART to fit business concurrency with periodic changes;

Step 3.2.1: Use the classification and regression tree CART to fit the business concurrency at {t ₁ , t ₂ , ..., t _n' } to obtain the fitted values {y(0), ... y(n'- 1), y(n')}, depicts the rising or falling trend of business concurrency;

Step 3.2.2: Compare the business concurrency obtained in step 3.2.1 with the real business concurrency to obtain the residual sequence {e(0), e(1),...,e(n)};

Step 3.2.3: Use the classification and regression tree CART to predict the business concurrency at the moment {t _n+1 , t _n+2 ..., t _m' } is {y(n+1), y(n+2), ..., y(m');

Step 3.2.4: Use the Fourier series FS to fit the residual sequence {e(0), e(1), ..., e(n)}, characterize the cyclical trend of business concurrency, and obtain { Residual value of business concurrency at time t _n'+1 , t _n'+2 ,..., t _m' } {e(n'+1), e(n'+2),...,e (m')};

Step 3.2.4.1: Use the function w(t) to fit the residual sequence e(0), e(1), ..., e(n'), the function w(t) is as follows:

表示向下取整，t＝1，2，...n’；Among them, a ₀ , a _j' and b _j' are all variables, P=n',

Indicates rounding down, t=1, 2,...n';

Step 3.2.4.2: Calculate the values of variables a _j' and b _j' by the least squares method, as shown in the following formulas:

Among them, w _j' is the j'th function used to fit the residual;

Step 3.2.5: Add the business concurrency at {t _n'+1 , t _n'+2 , ..., t _m' } to its corresponding residual value to obtain {t _n'+1 , t _n'+2 , ..., t _m' } Predicted value of business concurrency, namely {y(n'+1)+e(n'+1), y(n'+2)+e(n '+2), ..., y(m')+e(m')}.

The beneficial effects of adopting the above technical solutions are: the present invention provides a method for predicting the concurrent traffic volume of different types of virtual machines, which divides the traffic concurrent traffic volume into periodic, ascending, descending, quadratic and random Different types of business concurrency have different prediction methods. Classifying each business concurrency before forecasting can not only train the business concurrency model in a targeted manner, but also model the same type of business concurrency. Parameters can also be shared. The method of the present invention predicts the concurrency of each business of the virtual machine, which can provide a basis for the increase or decrease of the virtual machine in the next step, and at the same time help to accurately estimate the software aging status of the virtual machine, so as to improve the performance and reliability of the working virtual machine. sexual purpose.

Description of drawings

Fig. 1 is the example topology diagram of the air ticket online ordering system provided by the embodiment of the present invention;

2 is a flowchart of a method for predicting different types of service concurrency for virtual machines according to an embodiment of the present invention;

3 is a schematic diagram of a prediction result of a secondary service concurrency provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of a prediction result of the concurrency of a periodically rising type of service provided by an embodiment of the present invention.

In the figure, 1, client; 2, load balancing Nginx; 3, switch; 4, server; 5, business database MySQL.

Detailed ways

The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following examples are intended to illustrate the present invention, but not to limit the scope of the present invention.

In this embodiment, the online air ticket ordering system simulates a PC-side user application, builds the service system on the Dawning server, simulates real business concurrency scenarios by pressurizing the air ticket online ordering system, and collects different business concurrency data as an example , using a method for predicting different types of business concurrency of virtual machines of the present invention to predict the business concurrency. The example topology is shown in Figure 1. Client 1 uses LoadRunner software to generate concurrent business access. It can simulate a large number of users clicking on the air ticket ordering system page at the same time. After LoadRunner sends page requests, load balancing Nginx2 realizes the reception and processing of business requests. Assignment, and finally the server 4 installs Tomcat and deploys the air ticket online booking system, which is responsible for reading and writing the business database MySQL5, and processing the requests sent by LoadRunner.

A method for predicting business concurrency for different types of virtual machines, as shown in Figure 2, includes the following steps:

Step 1: Collect historical business concurrency of virtual machines and preprocess them. The specific methods are as follows:

Step 1.1: Scan the business concurrency of virtual machines for a period of time to find the missing points of business concurrency;

Step 1.2: Process the scanned missing points of business concurrency;

Step 1.2.1: For the case of missing individual sampling points, use the average value of the business concurrency in the previous cycle and the next cycle to fill in. The calculation formula for the missing business concurrency con(t) in the t-th time period of the virtual machine is as follows: shown:

Step 1.2.2: For samples missing more than 90%, discard all samples and set the value of business concurrency during this period to zero; for example, in 20 consecutive sampling cycles, only 2 cycles collect business concurrency value, or even if all the data is empty, it can be considered that the business concurrency collected during this period is unreliable and cannot be included in the historical sequence for prediction;

Step 1.3: Perform outlier adjustment for the extremely small samples with abnormal fluctuations in the collected business concurrency;

Step 1.3.1: Combine the quartiles to calculate the upper limit H and lower limit L of the normal value of the virtual machine business concurrency in time t, as shown in the following formula:

H=Q3+k*(Q3-Q1) (2)

L=Q1-k*(Q3-Q1) (3)

Among them, Q1 represents the lower quartile, that is, the 25th percentile of the ascending sequence of business concurrency in time t, and Q3 represents the upper quartile, that is, the percent of the ascending sequence of business concurrency in time t Seventy-five points, k is used to describe the abnormal degree of unreasonable sampling points, generally 1.5 and 3, representing moderate and extreme respectively;

Step 1.3.2: Determine whether the data of each sampling point is normal through the Tukey test method, and adjust the abnormal values;

If the data value of the sampling point is determined to be a sample of wrong business concurrency, the wrong value will be discarded first, and then supplemented by the mean value filling method;

If the data value of the sampling point is determined to be a sample of normal business concurrency, no adjustment will be made;

Step 1.4: Adjust the data interval for the business concurrency and CPU utilization data collected from the log database or management log, and merge the collected data in seconds, minutes or hours;

When modeling business concurrent access, the business concurrency sampled at 1-second intervals fluctuates frequently, and the trend does not change significantly, so it is impossible to mine the changing features, and over-sampling makes the model more computationally expensive and slower to train; therefore , in this embodiment, the average value is taken as an interval of 15 seconds to sort out the data, and other data of the virtual machine is also at an interval of 15 seconds;

Step 1.5: Use the maximum and minimum normalization method to normalize the data processed in step 1.4;

Step 2: Determine the type of virtual machine business concurrency based on the improved 1-Nearest Neighbor-Dynamic TimeWarping (1-NearestNeighbor-Dynamic TimeWarping, ie 1NN-DTW) method. The specific method is as follows:

Step 2.1: Classify the concurrency of each service of the virtual machine into rising, falling, quadratic, random, cyclically fluctuating, cyclically rising and cyclically falling;

Step 2.2: For various types of business concurrency, select the labeled business concurrency sequence as a known sample in advance;

Step 2.3: For each sequence of business concurrency to be classified, scan all known samples in turn and calculate the closest known sample through the proximity algorithm, then the type of the known sample is the type of business concurrency to be classified ;

Step 2.4: Classify all business concurrency into two categories to simplify the 1-nearest neighbor model;

Classify random, rising, falling, and quadratic business concurrency into categories that do not have periodic changes;

Classify the cyclically fluctuating, cyclically rising, and cyclically declining business concurrency as categories with cyclical changes;

Step 2.5: Construct an n×m matrix so that the sequence of business concurrency to be classified {x ₁ , x ₂ , ..., x _n } and a known sequence of business concurrency {y ₁ , y ₂ , ... , y _m } are aligned, where n is the total number of business concurrency to be classified, and m is the known total number of business concurrency;

Step 2.6: Take the two-point deviation of the i-th business concurrency x _i to be classified and the known j-th business concurrency amount y _j as the value d _{i, j} of the (i, j) position in the matrix, and use the Euclidean distance at the same time And the method of the square of the difference of the two-point derivative to calculate the sequence of business concurrency to be classified {x ₁ , x ₂ , ..., x _n } and the known business concurrency sequence {y ₁ , y ₂ , ... , y _m } The deviation d _{i, j} of each point after alignment is shown in the following formula:

d _i,j =(x _i -y _j ) ² +(x' _i -y' _j ) ² (4)

Among them, x′ _i and y′ _j are the derivatives of x _i and y _j respectively, and the estimation of the derivative x′ _i of the service concurrency x _i is shown in the following formula:

Step 2.7: Starting from position (1, 1) in the matrix, iteratively find a path with the smallest cumulative deviation according to the constraints that each position can only reach the position above, right or upper right except for the boundary value, until the end of position (n, m);

Step 3: Predict the business concurrency of different types of changes in the virtual machine. The specific methods are:

Step 3.1: Use a Classification and Regression Tree (CART) to fit the business concurrency without periodic changes;

Step 3.1.1: Traverse the arbitrary value f of each feature F of the sample business concurrency sequence, divide the sample data with (F, f) as the condition, determine the division position with the smallest square error, and select the highest value from the business concurrency sequence. good cutting point;

The calculation formula of the squared error error is as follows:

代表样本x中第i’个业务并发量的特征，y_i’代表分割前的第i’个序列样本，

代表分割后的第i’个子序列样本的拟合结果；in,

represents the feature of the i'th business concurrency in the sample x, y _i' represents the i'th sequence sample before splitting,

represents the fitting result of the i'th subsequence sample after segmentation;

Step 3.1.2: Save the business concurrency value as the cutting point, and perform segmentation on the business concurrency sequence;

Step 3.1.3: Construct subtrees with feature F greater than f and subtrees smaller than f in turn, and further iteratively segment and fit the sequence of business concurrency on the left and right of the current split point until it can no longer be divided into leaf nodes;

Step 3.1.4: Retraverse the sample data from bottom to top, check each split point for all business concurrency series, and judge the fitting error of the concurrency series before and after split,

If the fitting error of the concurrency sequence decreases after the split, the split point is retained;

If the fitting error of the concurrent series increases after the split, cancel the split point and merge the left and right series;

Step 3.2: Use Fourier series FS and classification and regression tree CART to fit business concurrency with periodic changes;

Step 3.2.1: Use the classification and regression tree CART to fit the business concurrency at {t ₁ , t ₂ , ..., t _n' } to obtain the fitted values {y(0), ... y(n'- 1), y(n')}, depicts the rising or falling trend of business concurrency;

Step 3.2.2: Compare the business concurrency obtained in step 3.2.1 with the real business concurrency to obtain the residual sequence {e(0), e(1),...,e(n)};

Step 3.2.3: Use the classification and regression tree CART to predict the business concurrency at {t _n+1 , t _n+2 , ..., t _m' } as {y(n+1), y(n+2) , ..., y(m');

Step 3.2.4: Use the Fourier series FS to fit the residual sequence {e(0), e(1), ..., e(n)}, characterize the cyclical trend of business concurrency, and obtain { Residual value of business concurrency at time t _n'+1 , t _n'+2 ,..., t _m' } {e(n'+1), e(n'+2),...,e (m')};

Step 3.2.4.1: Use the function w(t) to fit the residual sequence e(0), e(1), ..., e(n'), the function w(t) is as follows:

表示向下取整，t＝1，2，...n’；Among them, a ₀ , a _j' and b _j' are all variables, P=n',

Indicates rounding down, t=1, 2,...n';

Step 3.2.4.2: Calculate the values of variables a _j' and b _j' by the least squares method, as shown in the following formulas:

Among them, wi _' is the j'th function used to fit the residual;

Step 3.2.5: Add the business concurrency at {t _n'+1 , t _n'+2 , ..., t _m' } to its corresponding residual value to obtain {t _n'+1 , t _n'+2 , ..., t _m' } Predicted value of business concurrency, namely {y(n'+1)+e(n'+1), y(n'+2)+e(n '+2), ..., y(m')+e(m')}.

This embodiment also provides the use of the improved 1NN-DTW algorithm to determine the type of service concurrency, and compares it with the algorithm before the improvement to verify the accuracy of the improved 1NN-DTW, specifically:

First, use LoadRunner to record the access behavior of various services such as browsing, querying, and refunding of the server application. Then pressurize the server virtual machine for one hour and collect business concurrency, process missing and abnormal business concurrency values according to the preprocessing method, and adjust the concurrency data at 15-second intervals.

The improved 1NN-DTW algorithm is used to judge the type of concurrent traffic of the business, and compared with 1NN-DTW and 1NN-DDTW, the accuracy rate and the F value F-measure are used to measure the quality of each algorithm. The concurrency obtained in the first step is intercepted as a subsequence at every 80, 120, 160, and 200 sampling points, and the type labels are marked as a sample sequence according to the seven load variation trends listed in Table 1, and finally 700 420 sample sequences are selected as known samples for type judgment, and the remaining 280 samples are used as test samples.

Table 1 Concurrent access to different types of services

Table 2 shows the comparison results of the service concurrency type judgment using the improved 1NN-DTW of the present invention and the existing 1NN-DTW and 1NN-DDTW methods. It can be seen from Table 2 that the Accuracy and F-measure of the method of the present invention are significantly higher than those of the other two methods, indicating that when judging the type of business concurrency, the effect is better than the value and changing trend of the business concurrency. Just focus on one of them. In addition, although the method of the present invention calculates the Euclidean distance and the derivative difference of the similar points at the same time, the time used is not greatly increased.

Table 2 Classification of business concurrency by different methods

method Accuracy F-measure Time(ms) Improved 1NN-DTW 0.942 0.867 1120 1NN-DTW 0.873 0.751 984 1NN-DDTW 0.916 0.834 1097

This embodiment also provides using the method of the present invention to predict the service concurrency, and compares it with traditional methods such as ARIMA, specifically:

First, use LoadRunner to record the access behavior of various services such as browsing, querying, and refunding of the server application. Then pressurize the server virtual machine for one hour and collect the business concurrency, process the missing and abnormal business concurrency values according to the method described in the preprocessing, and adjust the concurrency data at 15-second intervals.

Two types of relatively complex concurrency, quadratic type and periodic rising type, are selected for prediction. By analyzing the business concurrency value in the past 25 minutes, the business concurrency amount in the next 5 minutes is estimated, and three evaluation criteria of mean square error (MSE), absolute error (MAE), and time duration are selected. With the help of Python toolkit, this method is combined with ARIMA and exponential smoothing. Holt-Winters comparison verifies the accuracy of the method of the present invention.

The business concurrency prediction results of the three methods are shown in Table 3, and the comparison results between the business concurrency prediction results of the three methods and the actual concurrency amount are shown in Figures 3 and 4. It can be seen from the figure that, compared with ARIMA and Holt-Winters, the method of the present invention fits the real business concurrency sequence better under the two set conditions, which shows that the method of the present invention can predict the concurrency of various types. more effective. According to further analysis of the results in Table 3, compared with ARIMA and Holt-Winters, the method of the present invention has the lowest MSE and MAE under the two types of concurrency scenarios. In the quadratic concurrency scenario, the MSE and MAE of the three methods are relatively close, but in the cyclically rising concurrency scenario, this method is significantly better. ARIMA and Holt-Winters have poor learning ability for such complex concurrency. These show that the method of the present invention has considerable accuracy in various scenarios.

Table 3 Business concurrency prediction results of different methods

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or some or all of the technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the scope defined by the claims of the present invention.

Claims (6)

1. a kind of business concurrency prediction method for different types of virtual machines, is characterized in that: comprise the following steps: Step 1: Collect historical business concurrency of virtual machines and preprocess them. The specific methods are as follows: Step 1.1: Scan the business concurrency of virtual machines for a period of time to find the missing points of business concurrency; Step 1.2: Process the scanned missing points of business concurrency; Step 1.3: Perform outlier adjustment for the extremely small samples with abnormal fluctuations in the collected business concurrency; Step 1.4: Adjust the data interval for the business concurrency and CPU utilization data collected from the log database or management log, and merge the collected data in seconds, minutes or hours; Step 1.5: Use the maximum and minimum normalization method to normalize the data processed in step 1.4; Step 2: Determine the type of virtual machine business concurrency based on the improved 1-nearest neighbor-dynamic time adjustment method 1NN-DTW. The specific method is: Step 2.1: Classify the concurrency of each service of the virtual machine into rising, falling, quadratic, random, cyclically fluctuating, cyclically rising and cyclically falling; Step 2.2: For various types of business concurrency, select the labeled business concurrency sequence as a known sample in advance; Step 2.3: For each sequence of business concurrency to be classified, scan all known samples in turn and calculate the closest known sample through the proximity algorithm, then the type of the known sample is the type of business concurrency to be classified ; Step 2.4: Classify all business concurrency into two categories to simplify the 1-nearest neighbor model; Classify random, rising, falling, and quadratic business concurrency into categories that do not have periodic changes; Classify the cyclically fluctuating, cyclically rising, and cyclically declining business concurrency as categories with cyclical changes; Step 2.5: Construct an n×m matrix so that the sequence of business concurrency to be classified {x ₁ ,x ₂ ,…,x _n } and a known sequence of business concurrency {y ₁ ,y ₂ ,…,y _m } Alignment, where n is the total number of business concurrency to be classified, and m is the known total number of business concurrency; Step 2.6: Take the two-point deviation of the i-th business concurrency x _i to be classified and the known j-th business concurrency amount y _j as the value d _{i, j} of the (i, j) position in the matrix, and use the Euclidean distance at the same time And the method of the square of the difference between the two-point derivatives to calculate the sequence of business concurrency to be classified {x ₁ ,x ₂ ,…,x _n } and the known business concurrency sequence {y ₁ ,y ₂ ,…,y _m } The deviation d _i,j of each point after alignment is shown in the following formula: d _i,j =(x _i -y _j ) ² +(x _i ′-y _j ′) ² (1) Among them, x _i ′ and y _j ′ are the derivatives of x _i and y _j respectively, and the estimation of the derivative x _i ′ of the service concurrency x _i is shown in the following formula:

Step 2.7: Starting from position (1,1) in the matrix, iteratively find a path with the smallest cumulative deviation according to the constraint that each position can only reach the position above, right or upper right except for the boundary value, until the end of position (n,m); Step 3: Predict the business concurrency of different types of changes in the virtual machine. The specific methods are: Step 3.1: Use the classification and regression tree CART to fit the business concurrency without periodic changes; Step 3.2: Use Fourier series FS and classification and regression tree CART to fit the business concurrency with periodic variation. 2. a kind of business concurrency prediction method for different types of virtual machines according to claim 1, is characterized in that: the concrete method of described step 1.2 is: Step 1.2.1: For the case of missing individual sampling points, use the average value of the business concurrency of the previous cycle and the next cycle to fill in. The calculation formula for the missing business concurrency con(t) of the virtual machine in the t-th time period is as follows: shown:

Step 1.2.2: For the case where the sample missing reaches more than 90%, discard all samples and set the value of the business concurrency in the scanning period to zero. 3. a kind of business concurrency prediction method for different types of virtual machines according to claim 1, is characterized in that: the concrete method of described step 1.3 is: Step 1.3.1: Combine the quartiles to calculate the upper limit H and lower limit L of the normal value of the virtual machine business concurrency in time t, as shown in the following formula: H=Q3+k*(Q3-Q1) (4) L=Q1-k*(Q3-Q1) (5) Among them, Q1 represents the lower quartile, that is, the 25th percentile of the ascending sequence of business concurrency in time t, and Q3 represents the upper quartile, that is, the percent of the ascending sequence of business concurrency in time t Seventy-five points, k is used to describe the abnormal degree of unreasonable sampling points, generally 1.5 and 3, representing moderate and extreme respectively; Step 1.3.2: Determine whether the data of each sampling point is normal through the Tukey test method, and adjust the abnormal values; If the data value of the sampling point is determined to be a sample of wrong business concurrency, the wrong value will be discarded first, and then supplemented by the mean value filling method; If the data value of the sampling point is determined to be a sample of normal business concurrency, no adjustment will be made. 4. a kind of business concurrency prediction method for different types of virtual machines according to claim 1, is characterized in that: the concrete method of described step 3.1 is: Step 3.1.1: Traverse the arbitrary value f of each feature F of the sample business concurrency sequence, divide the sample data with (F, f) as the condition, determine the segmentation position with the smallest square error, and select the highest value from the business concurrency sequence. good cutting point; The calculation formula of the squared error error is as follows:

in,

represents the feature of the i'th business concurrency in the sample x, y _i' represents the i'th sequence sample before splitting,

represents the fitting result of the i'th subsequence sample after segmentation; Step 3.1.2: Save the business concurrency value as the cutting point, and perform segmentation on the business concurrency sequence; Step 3.1.3: Construct subtrees with feature F greater than f and subtrees smaller than f in turn, and further iteratively segment and fit the sequence of business concurrency on the left and right of the current split point until it can no longer be divided into leaf nodes; Step 3.1.4: Retraverse the sample data from bottom to top, check each split point for all business concurrency series, and judge the fitting error of the concurrency series before and after split, If the fitting error of the concurrency sequence decreases after the split, the split point is retained; If the fitting error of the concurrent series increases after the split, cancel the split point and merge the left and right series. 5. a kind of business concurrency prediction method for different types of virtual machines according to claim 4, is characterized in that: the concrete method of described step 3.2 is: Step 3.2.1: Use the classification and regression tree CART to fit the business concurrency at {t ₁ ,t ₂ ,...,t _n' } to obtain the fitted value {y(0),...y(n'-1),y (n')}, depicting the rising or falling trend of business concurrency; Step 3.2.2: Compare the business concurrency obtained in step 3.2.1 with the real business concurrency to obtain the residual sequence {e(0),e(1),...,e(n)}; Step 3.2.3: Use the classification and regression tree CART to predict the business concurrency at the moment {t _n+1 ,t _n+2 ,...,t _m' } as {y(n+1), y(n+2),... ,y(m'); Step 3.2.4: Use Fourier series FS to fit the residual sequence {e(0),e(1),...,e(n)}, characterize the periodic trend of business concurrency, and obtain {t _{n '+1} ,t _n'+2 ,...,t _m' } Residual value of business concurrency at moment {e(n'+1),e(n'+2),...,e(m')}; Step 3.2.5: Add the business concurrency at {t _n'+1 ,t _n'+2 ,...,t _m' } to its corresponding residual value to obtain {t _n'+1 ,t _{n' +2} ,...,t _m' } Predicted value of business concurrency at the moment, namely {y(n'+1)+e(n'+1), y(n'+2)+e(n'+2) ,…,y(m')+e(m')}. 6. a kind of business concurrency prediction method for different types of virtual machines according to claim 5, is characterized in that: the concrete method of described step 3.2.4 is: Step 3.2.4.1: Use the function w(t) to fit the residual sequence e(0), e(1),...,e(n'), the function w(t) is as follows:

Among them, a ₀ , a _j' and b _j' are all variables, P=n',

Indicates rounding down, t=1,2,...n'; Step 3.2.4.2: Calculate the values of variables a _j' and b _j' by the least squares method, as shown in the following formulas:

where w _j' is the j'th function used to fit the residuals.

Images