WO2023024679A1 - Method and apparatus for predicting server capacity - Google Patents

Abstract

Provided in the embodiments of the present invention are a method and apparatus for predicting server capacity. The method comprises: for any statistical dimension in a service system, analyzing a historical activity degree of the statistical dimension in each historical time period, and determining a law feature that complies with the historical activity degree of the statistical dimension in each historical time period; obtaining a superimposed predicted TPS of the service system in each prediction time period by means of the law feature of each statistical dimension; converting the superimposed predicted TPS of the service system in each prediction time period into a trusted predicted TPS of the service system in each prediction time period according to a conversion relationship; and determining the server capacity of the service system in each prediction time period by means of the trusted predicted TPS of the service system in each prediction time period. In this way, trusted predicted TPSs that are obtained can more accurately reflect a change law of different statistical dimensions in prediction time periods, thereby accurately predicting the server capacity of a service system in each prediction time period.

Description

A method and device for predicting server capacity

Cross References to Related Applications

This application claims the priority of the Chinese patent application submitted to the China Patent Office on August 27, 2021, with the application number 202110992903.2 and the application name "A Method and Device for Predicting Server Capacity", the entire contents of which are incorporated herein by reference. Applying.

technical field

The embodiments of the present invention relate to the technical field of financial technology, and in particular to a method, device, computing device, and computer-readable storage medium for predicting server capacity.

Background technique

With the development of computer technology, more and more technologies are applied in the financial field, and the traditional financial industry is gradually transforming into Fintech. However, due to the security and real-time requirements of the financial industry, higher requirements are placed on technology. requirements.

With the rapid development and progress of Internet technology, the number of Internet users has increased dramatically. Therefore, a system often needs to deploy multiple servers to meet the transaction volume requested by Internet users. If the deployed server capacity (that is, the number of servers) is too small, it is easy for the transaction volume of requests to exceed the rated load limit of the server, resulting in server congestion or even collapse; if the deployed server capacity is too large, it will not only cause waste of computing resources, but also Also increased the unnecessary cost of the enterprise.

At present, the server capacity is manually adjusted by manually observing the server load changes or monitoring system alarms (network congestion or crash of the server), which easily leads to misoperation and is less intelligent.

Therefore, how to predict the capacity of servers in advance and use the predicted results to guide future resource deployment is very important for the business development of information systems.

To sum up, the embodiments of the present invention provide a method for predicting server capacity, which is used to improve the accuracy of server capacity prediction, so as to flexibly adjust the server capacity.

Contents of the invention

An embodiment of the present invention provides a method for predicting server capacity, which is used to improve the accuracy of server capacity prediction, thereby flexibly adjusting the server capacity.

In a first aspect, an embodiment of the present invention provides a method for predicting server capacity, including:

For any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period;

According to the regular characteristics of each statistical dimension, predict the predicted transaction volume per second TPS of each statistical dimension in each forecast period, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the superposition forecast of the business system in each forecast period TPS;

Convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the TPS of the business system in each historical period The relationship between the superimposed historical TPS and the credible historical TPS is obtained by analyzing the relationship; the superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; the credible historical TPS is obtained by analyzing the business system in It is obtained by analyzing the TPS of each historical period;

The server capacity of the service system in each forecast period is determined through the credible predicted TPS of the service system in each forecast period.

Since user activity will change in different historical periods, and in different statistical dimensions, the change rules are different, so by analyzing the historical activity of each historical period in different statistical dimensions in the business system, we can get The regular characteristics of the historical activity of each statistical dimension in each historical period can more accurately predict the predicted TPS of each statistical dimension in each forecast period. Then, by superimposing the predicted TPS of different statistical dimensions, the superimposed predicted TPS of the entire business system can be further determined. The superimposed predicted TPS obtained in this way can more accurately reflect the changing rules of different statistical dimensions in each forecast period. In order to further improve the accuracy of the forecast, through the analysis of the credible historical TPS and superimposed historical TPS of the business system in each historical period, the conversion relationship between the two is determined, and the credible predicted TPS is obtained according to the conversion relationship and the superimposed predicted TPS. The credible predicted TPS is more accurate. Therefore, the server capacity of the business system in each forecast period can be accurately predicted.

Optionally, the historical activity of any statistical dimension in each historical period is obtained by the following methods, including:

The quotient of the historical TPS of the statistical dimension in each historical period and the number of first accounts in each historical period is determined as the historical activity of the statistical dimension in each historical period.

Optionally, determining the regular characteristics conforming to the historical activity of the statistical dimension in each historical period includes:

Fitting the historical activity of the statistical dimension in each historical period to obtain a fitting curve corresponding to each preset function;

A fitting curve corresponding to each preset function conforming to a preset condition is determined as a regular feature. Optionally, determining the regular characteristics conforming to the historical activity of the statistical dimension in each historical period includes:

If it is determined that there is no fitting curve that meets the preset conditions, the historical activity of the statistical dimension in each historical period is segmented into intervals;

For any interval segment, the floating coefficient of each historical activity in the interval segment is determined as the regular feature of the interval segment; the floating coefficient is determined according to the floating ratio of each historical activity; The floating ratio is the ratio of the difference between the highest value of historical activity and the lowest value of historical activity in the interval segment to the lowest value of historical activity.

Optionally, predict the predicted TPS of each statistical dimension in each prediction period through the regular characteristics of each statistical dimension, including:

Determine the forecast activity of each statistical dimension in each forecast period through the regular characteristics of each statistical dimension;

The product of the forecast activity of each statistical dimension in each forecast period and the second account number in the same forecast period is determined as the predicted TPS of each statistical dimension in each forecast period.

Optionally, determining the server capacity of the business system in each forecast period through the trusted predicted TPS of the business system in each forecast period, including:

Collect the server performance indicators of the business system in each historical period, and determine the TPS carried by the server when the server performance indicators meet the set conditions as the stand-alone TPS of the server;

The server capacity of the service system in each forecast period is determined through the trusted forecast TPS and the stand-alone TPS.

Optionally, determining the server capacity of the business system in each prediction period through the trusted predicted TPS and the stand-alone TPS includes:

The first range of server capacity is determined according to the following formula:

Wherein, T _max is the trusted predicted TPS, N is the server capacity, s is the maximum failure time of a single physical machine, n is the number of servers deployed on a single physical machine, and B is the transaction loss threshold _once ;

The server capacity is determined within the first range of server capacities.

Optionally, determining the server capacity of the business system in each prediction period through the trusted predicted TPS and the stand-alone TPS includes:

The second range of server capacity is determined according to the following formula:

Wherein, the L _{computer room} is the number of computer rooms of the business system, and the T _stand-alone is the single-machine TPS;

The server capacity is determined within the first range of server capacities and the second range of server capacities.

Optionally, determining the server capacity of the business system in each prediction period through the trusted predicted TPS and the stand-alone TPS includes:

The third range of server capacity is determined according to the following formula:

Wherein, ∈ is the highest load rate;

The server capacity is determined within the first range of server capacities and the third range of server capacities.

In the second aspect, the embodiment of the present invention also provides an apparatus for predicting server capacity, including:

Identify units for:

For any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period;

According to the regular characteristics of each statistical dimension, predict the predicted transaction volume per second TPS of each statistical dimension in each forecast period, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the superposition forecast of the business system in each forecast period TPS;

Convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the TPS of the business system in each historical period The relationship between the superimposed historical TPS and the credible historical TPS is obtained by analyzing the relationship; the superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; the credible historical TPS is obtained by analyzing the business system in It is obtained by analyzing the TPS of each historical period;

The server capacity of the service system in each forecast period is determined through the credible predicted TPS of the service system in each forecast period.

Optionally, the determining unit is specifically configured to:

The quotient of the historical TPS of the statistical dimension in each historical period and the number of first accounts in each historical period is determined as the historical activity of the statistical dimension in each historical period.

Optionally, the determining unit is specifically configured to:

Fitting the historical activity of the statistical dimension in each historical period to obtain a fitting curve corresponding to each preset function;

A fitting curve corresponding to each preset function conforming to a preset condition is determined as a regular feature.

Optionally, the determining unit is specifically configured to:

If it is determined that there is no fitting curve that meets the preset conditions, the historical activity of the statistical dimension in each historical period is segmented into intervals;

For any interval segment, the floating coefficient of each historical activity in the interval segment is determined as the regular feature of the interval segment; the floating coefficient is determined according to the floating ratio of each historical activity; The floating ratio is the ratio of the difference between the highest value of historical activity and the lowest value of historical activity in the interval segment to the lowest value of historical activity.

Optionally, the determining unit is specifically configured to:

Determine the forecast activity of each statistical dimension in each forecast period through the regular characteristics of each statistical dimension;

The product of the forecast activity of each statistical dimension in each forecast period and the second account number in the same forecast period is determined as the predicted TPS of each statistical dimension in each forecast period.

Optionally, the determining unit is specifically configured to:

Collect the server performance indicators of the business system in each historical period, and determine the TPS carried by the server when the server performance indicators meet the set conditions as the stand-alone TPS of the server;

The server capacity of the service system in each forecast period is determined through the trusted forecast TPS and the stand-alone TPS.

Optionally, the determining unit is specifically configured to:

The first range of server capacity is determined according to the following formula:

Wherein, T _max is the trusted predicted TPS, N is the server capacity, s is the maximum failure time of a single physical machine, n is the number of servers deployed on a single physical machine, and B is the transaction loss threshold _once ;

The server capacity is determined within the first range of server capacities.

Optionally, the determining unit is specifically configured to:

The second range of server capacity is determined according to the following formula:

Wherein, the L _{computer room} is the number of computer rooms of the business system, and the T _stand-alone is the single-machine TPS;

The server capacity is determined within the first range of server capacities and the second range of server capacities.

Optionally, the determining unit is specifically configured to:

The third range of server capacity is determined according to the following formula:

Wherein, ∈ is the highest load rate;

The server capacity is determined within the first range of server capacities and the third range of server capacities.

In a third aspect, an embodiment of the present invention also provides a computing device, including:

memory for storing computer programs;

The processor is configured to call the computer program stored in the memory, and execute the method for predicting server capacity listed in any of the above-mentioned ways according to the obtained program.

In the fourth aspect, the embodiment of the present invention also provides a computer-readable storage medium, the computer-readable storage medium stores a computer-executable program, and the computer-executable program is used to make the computer execute any of the methods listed above. A method for predicting server capacity.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

FIG. 1 is a method for predicting server capacity provided by an embodiment of the present invention;

Fig. 2A is a schematic diagram of the historical activity of a possible Taobao account related to withdrawing money provided by an embodiment of the present invention;

Fig. 2B is a schematic diagram of the historical activity of a possible Taobao account related to the withdrawal of funds provided by the embodiment of the present invention;

Fig. 3A is a schematic diagram of the daily consumption day historical activity of a Taobao account in each historical period of withdrawal provided by an embodiment of the present invention;

Fig. 3B is a schematic diagram of the predicted activity of a Taobao account on the daily consumption day of each forecast period for withdrawal provided by an embodiment of the present invention;

FIG. 3C is a schematic diagram of the historical activity of a Taobao account on special holidays in various historical periods of withdrawal provided by an embodiment of the present invention;

Fig. 3D is a schematic diagram of the prediction activity of a Taobao account on special holidays in each prediction time period for withdrawing money provided by an embodiment of the present invention;

FIG. 3E is a schematic diagram of the predicted activity of a Taobao account in each predicted time period for withdrawing money provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of a possible conversion relationship K provided by an embodiment of the present invention;

FIG. 5 is a schematic diagram of a trusted predicted TPS for each prediction period provided by an embodiment of the present invention;

Fig. 6 is a possible schematic diagram of the server capacity of the business system in each month of the next year provided by the embodiment of the present invention;

Fig. 7 shows a possible computer room environment schematic diagram of the business system;

FIG. 8 is a schematic structural diagram of an apparatus for predicting server capacity provided by an embodiment of the present invention;

FIG. 9 is a schematic structural diagram of a computer device provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose, implementation and advantages of the application clearer, the following will clearly and completely describe the exemplary embodiments of the application in conjunction with the accompanying drawings in the exemplary embodiments of the application. Obviously, the described exemplary embodiments It is only a part of the embodiments of the present application, but not all the embodiments.

Based on the exemplary embodiments described in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts fall within the protection scope of the appended claims of this application. In addition, although the disclosures in this application are introduced as exemplary one or several examples, it should be understood that each aspect of these disclosures can also independently constitute a complete implementation.

It should be noted that the brief description of the terms in this application is only for the convenience of understanding the implementations described below, and is not intended to limit the implementations of this application. These terms are to be understood according to their ordinary and usual meaning unless otherwise stated.

The terms "first", "second", and "third" in the description and claims of this application and the above drawings are used to distinguish similar or similar objects or entities, and do not necessarily mean limiting specific Sequential or sequential order, unless otherwise indicated (Unless otherwise indicated). It should be understood that the terms used in this way can be interchanged under appropriate circumstances, for example, they can be implemented in a sequence other than those shown in the illustrations or descriptions of the embodiments of the present application.

Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover but not exclusively include, for example, a product or device comprising a series of components need not be limited to those components explicitly listed, but may include other components not expressly listed or inherent in these products or equipment.

Figure 1 exemplarily shows a method for predicting server capacity provided by an embodiment of the present invention, including the following steps:

Step 101 , for any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period.

Step 102, predict the predicted TPS (Transaction Per Second, transaction volume per second) of each statistical dimension in each forecast period according to the regular characteristics of each statistical dimension, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the business system Overlay predicted TPS at each forecast period.

Step 103, convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the superimposed historical TPS and predictable TPS of the business system in each historical period It is obtained by analyzing the relationship between trusted and historical TPS; superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; trusted historical TPS is obtained by analyzing the TPS of the business system in each historical period.

Step 104, determine the server capacity of the business system in each forecast period according to the credible predicted TPS of the business system in each forecast period.

The method of the embodiment of the present invention can predict the future TPS change trend according to the historical TPS change trend, for example, analyze the TPS change trend of this month and predict the TPS change trend of next month; it can also analyze the TPS change trend of this year , predict the changing trend of TPS next year; you can also analyze the changing trend of TPS in recent years, and predict the changing trend of TPS next year or in the next few years. The embodiments of the present invention do not limit this. The embodiment of the present invention is introduced by taking the prediction of the TPS change trend of next year according to the TPS change trend of this year as an example.

In step 101, any statistical dimension in the business system is first analyzed to determine the historical activity of the statistical dimension in each historical period.

In fact, the historical activity will change over time. For example, for the payment system connecting the e-commerce platform and the bank’s core system, the e-commerce platform will carry out commercial promotions every special holiday, leading to a sudden increase in historical activity. This leads to a higher TPS required by the payment system under the same number of valid accounts. Therefore, we need to conduct a detailed analysis of the historical activity of the data throughout the year, and find the rules in sections.

Optionally, the historical activity of each day in a year can be analyzed to determine the predicted activity of each day in the next year; the historical activity of each month in a year can also be analyzed to determine the predicted activity in the next year The forecast activity of each month; the historical activity of each quarter of the year can also be analyzed to determine the forecast activity of each quarter in the next year. Those skilled in the art can make various choices according to needs, which are not limited here. In the embodiment of the present invention, the analysis is made based on the historical activity of each month in a year, and the predicted activity of each month in the next year is obtained as an example, and the introduction is made.

At the same time, since there are many statistical dimensions of the business system, in different statistical dimensions, the change rules of historical activity in each historical period are different. The statistical dimension may be account type, transaction type, etc., which are not limited in this embodiment of the present invention. If the statistical dimension is account type, it can be found that different account types and users have different usage habits, so the change rules of historical activity are different. For example, for e-commerce platforms, the daily historical activity and holiday historical activity will be different There is a big difference, because during special holidays, e-commerce companies will carry out preferential marketing and promotion, which makes the activity of existing users increase sharply; while other businesses with fixed models such as water and electricity bill payment, the historical activity will basically not change due to holidays And change. If the statistical dimension is the transaction type, the transaction type can be withdrawal, deposit, query, etc. For these different transaction types, the user's usage habits are also different, and correspondingly, the change law of historical activity will also be quite different .

In order to be able to take into account the changing rules of the historical activity of each historical period in different statistical dimensions, the embodiment of the present invention aims at any statistical dimension in the business system to determine the historical activity of the statistical dimension in each historical period, for example, for the account When the type is a Taobao account, determine the historical activity of each month of the year, or for the transaction type of withdrawal, determine the historical activity of each month of the year, or for the account type of Taobao account , respectively determine the historical activity of each month of this year when the transaction type is withdrawal and deposit. The foregoing is only an example, and this embodiment of the present invention does not limit it.

Collect the TPS of the whole year in the business system. TPS mainly adopts the method of log collection. The TPS in the business system is obtained through log analysis. The number of valid accounts is collected by the database. The number of valid accounts at the end of each day is the effective number of accounts.

Optionally, the data can be screened when collecting the annual TPS in the business system. A business system may have a sudden increase or decrease in the overall transaction volume throughout the year due to internal system machine failures, system internal production pressure testing, related party failures, etc., thereby affecting historical activity. Our model calculation must be carried out when the account is naturally active, otherwise it will affect the overall reliability. This requires removing these glitches. Only after removing these glitches can we get the historical activity under the overall natural activity throughout the year. For example, it can be screened according to server performance indicators. By monitoring server performance indicators such as CPU usage and memory usage, it is determined that the time when the server performance indicators are abnormal is the time of machine failure, and then the TPS at the time of failure is eliminated; it can also be determined by Manually enter the failure time, such as the internal production pressure test of the system, related party failure and other unconventional time periods, which can be eliminated by manually entering the time point.

For the statistical dimension where the account type is Taobao account and the transaction type is gold withdrawal, the historical TPS of Taobao account withdrawals in each month of the year is obtained. For example, there are 31 days in January, 24 hours per day, and 60 minutes, 60s per minute, then for each day, you can get 24*60*60 TPS, determine the highest TPS among these TPS as the highest TPS of the day, and then determine the highest TPS of the month among the 31 highest TPS of the day as Historical TPS in January; similarly, historical TPS in February, March...December. Obtain the number of valid accounts in the current month, for example, the number of valid accounts on the last day, or the highest number of accounts in the current month. The number of valid accounts in the current month is taken as the first number of accounts. According to the historical TPS of each month and the number of the first account of each month, determine the historical activity of the Taobao account in each month of the year, specifically, the historical TPS of each month divided by each month The number of the first account of , get the historical activity of each month. FIG. 2A shows a schematic diagram of a possible monthly historical activity of a Taobao account related to gold withdrawal.

For the statistical dimension that the account type is Taobao account and the transaction type is gold deposit, the historical TPS related to the deposit of Taobao account in each month of this year is obtained, and combined with the number of first accounts in each month, we can get the Taobao account in this year. The specific implementation method of the monthly historical activity related to the deposit is the same as the method of determining the historical activity related to the withdrawal of the Taobao account in each month, and will not be repeated here.

For the statistical dimension that the account type is a Meituan account and the transaction type is withdrawal, the historical TPS of the withdrawal of the Meituan account in each month of the year is obtained, and combined with the number of first accounts in each month, the Meituan account can be obtained. The specific implementation method of the historical activity related to gold withdrawal in each month of this year is the same as the method of determining the historical activity related to gold withdrawal in each month of the Taobao account, and will not be repeated here.

For the statistical dimension that the account type is a Meituan account and the transaction type is a deposit, the historical TPS of the Meituan account for each month of this year is obtained. The specific implementation method of the historical activity related to the deposit in each month of the year is the same as the method of determining the historical activity related to the withdrawal of the Taobao account in each month, and will not be repeated here.

So far, the historical activity of the four statistical dimensions in each historical period in the business system has been obtained. The change trend of the historical activity of each statistical dimension in each historical period is different, so it needs to be analyzed separately.

For the statistical dimension where the account type is Taobao account and the transaction type is gold withdrawal, Figure 2A shows a possible schematic diagram of the monthly historical activity of Taobao account related to gold withdrawal. The historical activity data of 12 months is a A group of scattered point data, according to the group of scattered point data, determine the regular characteristics of the historical activity of each historical period in the statistical dimension. Then, according to the regular characteristics, the forecast activity in each forecast period is determined, that is, the forecast activity in the next 12 months is determined by analyzing the historical activity in the 12 months of this year.

Specifically, the embodiment of the present invention provides two methods for determining the forecast activity in each forecast period.

method one

The historical activity of each historical period is fitted through a series of curve models, such as linear models, logarithmic models, power function models, or exponential models.

(1) Fitting by linear model

Fitting can be performed, for example, by the method of least squares.

The formula for the linear equation is:

y=a+bx (1)

In this formula, a and b are arbitrary real numbers, x is the timeline, and y is the historical activity we want to obtain. If we want to obtain this unary linear equation, we must determine the values of the real numbers a and b. Set the actual historical activity value as yi, and the approximate value obtained by using the formula is Yi. We hope that the smaller the difference between the historical activity calculated by the formula and the actual historical activity, the better, that is to say Our process of finding the linear regression equation is actually the process of finding the minimum value of the dispersion yi-Yi, but the value of the dispersion can be positive or negative, and direct addition will cancel out the true closeness of the data, so this dispersion It cannot be directly obtained by summing the subtracted values, as shown in the following formula 2:

Instead, the minimum sum of squares of the dispersion should be the basis for optimization, as shown in Formula 3 below:

Put formula 1 into formula 3 to get formula 4:

When the sum of squared deviations Q ² reaches the minimum, the straight line obtained by regression is the straight line with the smallest deviation from our actual data as a whole.

公式5： Next, we need to set the mean value of x as

Formula 5:

公式6： The average value of y is

Formula 6:

According to formula 4, formula 7 is obtained:

Next, decompose and simplify the above formula

①

Solving process:

②

Solving process:

Therefore, according to the two formulas ① and ②, formula 7 can be simplified

与a、b无关，属于常数项，想要让R ²最小，那么就必须让前面与a、b相关的多项式尽量小，最好是等于0，即 So far we can get the last two

It has nothing to do with a and b and is a constant term. If you want to minimize R ² , you must make the previous polynomials related to a and b as small as possible, preferably equal to 0, that is

③

④

from ③

From ④

Therefore, the values of the linear equations a and b can be obtained according to the known values of all scattered points ( _xi , y _i ), so as to obtain the linear regression equation y=a+bx.

(2) Logarithmic model

The logarithmic function is defined as follows

y=a+blnx

Among them, lnx represents the logarithm of the natural number of x, which is nonlinear. In order to simplify the calculation, we can linearize the nonlinear model of the logarithm

Let x ^* = lnx

Then for the variables y and x ^* , that is, y and lnx have been transformed into a linear relationship, where a and b are the coefficients to be obtained.

After converting into a linear relationship, the values of a and b can be obtained through the linear regression algorithm.

(3) Power function model

The power function is defined as follows

y=ax ^b

Similarly, a and b are the constants we need to obtain, and the relationship between x and y is also nonlinear. In order to simplify the calculation, take the logarithm of the natural numbers on both sides of the formula at the same time to get

lny=lna+blnx

Let y ^* =lny, a ^* =lna, x ^* =lnx

The formula can be converted into y ^* = a ^* + bx ^* , where a ^* and b are parameters to be estimated, and x ^* and y ^* are linearly related. The values and calculation formulas of a and b can be obtained according to linear regression.

(4) Exponential function model

The exponential function is defined as follows

y=ae ^bx

Among them, a and b are constants, and the relationship between x and y is also nonlinear. Take the logarithm of the natural numbers on both sides of the formula to get

lny=lna+bx

Let y ^* =lny, a ^* =lna, then

y ^* = a ^* + bx

Among them, a ^* and b are the parameters to be estimated, and x and y ^* are in a linear relationship, and the result value can be obtained according to the linear regression.

为所有实际散点y _i的平均值。那么， The historical activity of each historical period is fitted through the above series of curve models. Different curve models get different a and b result values. Here we need to further analyze the results to determine whether the curve we get is close to For the given scatter data, y _i is the actual value of the i scatter data, and Y _i is the i value calculated according to the obtained formula,

is the average value of all actual scatter points y _i . So,

total sum of squares

explain sum of squares

coefficient of determination

When the determination coefficient R ² =1, it becomes a perfect fit, which means that the data calculated by the formula completely coincides with the scatter data we actually give, but this is difficult to achieve, because there are certain errors in things in nature, when When the coefficient of determination R ² →1, it means that the formula obtained by our calculation model is more accurate. Therefore, after obtaining the model formula based on the above several curve models, the R ² value of each model can be calculated. The closer the R ² value is to 1, the more fitting curve we want to obtain. The fitting curve that meets the set conditions is the predicted activity of the Taobao account for each month related to the withdrawal of funds in the next year.

The predicted activity of each month of next year and the historical activity of each month of this year obtained through the above method will not be exactly the same, but are obtained according to the change trend of historical activity, which can more scientifically reflect the change law of predicted activity .

way two

In practice, most of the historical activity trends may not be the standard fitting curve as mentioned in Method 1, that is to say, the R ² value we calculated is far from 1, then The historical activity of the statistical dimension in each historical period can be divided into intervals and analyzed separately.

For example, divide historical activity into different time periods according to time, for example, divide 12 months into 4 intervals, and perform curve fitting on each interval through method 1.

For example, interval segmentation can also be performed according to daily consumption days and special holidays. Figure 2B shows a possible change trend of the annual historical activity of Taobao accounts for withdrawals. It can be seen that, in addition to several special events such as Spring Festival, Mid-Autumn Festival, New Year’s Day, 214 Valentine’s Day, and 520 Valentine’s Day, etc. Except for holidays, the data trend of the withdrawal account of this account type is basically the same, that is to say, the historical activity during the daily time period is basically maintained within a stable range; during the holiday time period, the historical activity is basically maintained within a stable range . This requires separating the historical activity of daily consumption days and special holidays for model calculation.

Extract the daily consumption days and the corresponding historical activity in Figure 2B, and get Figure 3A. It can be seen that for Taobao accounts, the historical activity related to withdrawals fluctuates between 0.5835-0.7886, and the floating ratio is determined =( Highest historical activity - lowest historical activity)/lowest historical activity, then in this example, the floating ratio is 35.15%. Then determine the floating coefficient according to the floating ratio as the prediction mode, and determine the floating coefficient a1 as (floating range+1), in this example, the floating coefficient is 1.3515.

Then multiply the historical activity of this year in Figure 3A by the floating coefficient a1 to get the predicted activity of the Taobao account and the daily consumption day related to withdrawal next year, as shown in Figure 3B.

Extract the special holidays and the corresponding historical activity in Figure 2B, and get Figure 3C. It can be seen that for Taobao accounts, the historical activity related to withdrawals fluctuates between 0.9-3.48, and the fluctuation is determined in the same way. Coefficient b1=2.87.

Then multiply the historical activity of each month of this year in Figure 3C by the floating coefficient b1 to get the predicted activity of the Taobao account and the special holidays related to cash withdrawal next year, as shown in Figure 3D.

Combining Figure 3B and Figure 3D, we can get the predicted activity of Taobao account related to withdrawal in the next year, as shown in Figure 3E.

The above method fully takes into account the user's consumption habits on daily consumption days and special holidays, and through targeted analysis of the user's consumption habits, it is possible to more accurately determine the predicted activity of the next year.

The above are the two methods for determining the forecast activity given by the embodiment of the present invention. According to the above methods, other statistical dimensions can be determined respectively, such as account type is Taobao + transaction type is deposit, account type is Meituan + transaction type is withdrawal Gold, the account type is Meituan + the transaction type is the forecast activity of deposit next year.

After determining the forecast activity of each statistical dimension for next year, the predicted TPS of each statistical dimension in each forecast period is determined according to the second account number of each statistical dimension in each forecast period.

The number of second accounts is input by business personnel according to the business promotion needs of next year, and the business personnel will input the number of second accounts in each forecast period next year into the model. For example, for Taobao accounts, the number of second accounts in each forecast period is 10,000 and 20,000 . Then, for the four different statistical dimensions determined above, the predicted TPS can be respectively determined in combination with the second account number.

For example, for the predicted activity of the Taobao account related to gold withdrawal, the obtained predicted activity of each forecast period is multiplied by the second account number (10,000, 20,000, 20,000...) of each corresponding forecast period, You can get the predicted TPS of Taobao account for each forecast period related to withdrawal. Similarly, for the predicted activity of Taobao account related to the deposit, the obtained predicted activity of each forecast period is multiplied by the corresponding number of second accounts (10,000, 20,000, 20,000...) in each forecast period, You can get the predicted TPS of Taobao account for each forecast period related to the deposit. For the predicted activity of the Meituan account related to withdrawal, the predicted activity of each forecast period obtained is multiplied by the number of second accounts (30,000, 10,000, 20,000...) corresponding to each forecast period, which can be obtained The predicted TPS of each forecast period for the withdrawal of the Meituan account. For the predicted activity of the Meituan account related to the deposit, multiply the obtained predicted activity of each forecast period by the number of second accounts (30,000, 10,000, 20,000...) corresponding to each forecast period to get The forecasted TPS of each forecast period for the deposit of the Meituan account.

So far, the predicted TPS of the four statistical dimensions in each forecast period has been obtained. Then, the predicted TPS of each statistical dimension in each forecast period is superimposed according to the same forecast period, and the superimposed predicted TPS of the business system in each forecast period can be obtained. The capacity of the server can be directly determined by dividing the superimposed predicted TPS by the loadable TPS of a single server, that is, the single-machine TPS.

However, looking back at the determination process of the superimposition predicted TPS, we can find that the superimposed predicted TPS is too large. Because we distinguish between different statistical dimensions, and in each statistical dimension, the highest TPS of the month is selected as the historical TPS of the statistical dimension, and when the business system is running normally, the highest TPS of the month for these four statistical dimensions It is often not generated at the same time, which will cause the value of the superimposed predicted TPS to be too large. For example, the Meituan account may reach a maximum TPS of 5 at 17 o'clock, and the Taobao account may achieve a maximum TPS of 6 at 2 o'clock, but the entire business system reaches a maximum TPS of 10 at 18 o'clock, then simply set the highest TPS of the Meituan account The 11 obtained by superimposing TPS and the highest TPS of the Taobao account is relatively large compared to the highest TPS of the business system which is 10. Therefore, we need to further correct the value of the superimposed predicted TPS to obtain a credible predicted TPS.

We know that there must be a conversion relationship between the superimposed forecast TPS and the credible forecast TPS, and this conversion relationship can be obtained by analyzing historical data.

Specifically, firstly, the historical TPS of each statistical dimension in each historical period is superimposed to obtain the superimposed historical TPS of each historical period. Then determine the credible historical TPS of the business system (that is, without distinguishing statistical dimensions) in each historical period, and determine the conversion relationship K according to the superimposed historical TPS and credible historical TPS in each historical period.

For example, first determine the historical TPS1 of the statistical dimension of the withdrawal of Taobao account in each historical period of this year, the historical TPS2 of the statistical dimension of deposit of Taobao account in each historical period of this year, and the statistical dimension of withdrawal of Meituan account In the historical TPS3 of each historical period of this year, the historical TPS4 of the statistical dimension of Meituan account deposits in each historical period of this year, the historical TPS of each historical period of the 4 statistical dimensions are superimposed according to the same historical period, and the superposition is obtained Historical TPS. Then, without distinguishing the statistical dimension, determine the credible historical TPS of the business system in each historical period of this year. For each historical period, divide the trusted historical TPS by the superimposed historical TPS to obtain the conversion relationship K. Figure 4 shows a schematic diagram of a possible conversion relationship K. As shown in the figure, each historical period will There is a corresponding conversion relationship K between superimposed historical TPS and credible historical TPS, that is, the ordinate in Fig. 4 .

We believe that the conversion relationship obtained from the analysis of historical data can still be applied to the results obtained from the analysis of future data, that is, the conversion relationship between the superimposed predicted TPS and the credible predicted TPS is still satisfied.

It can be seen that the larger K is, the larger the value of the trusted predicted TPS will be. In order to ensure that the number of servers deployed during the prediction period is large enough to prevent the system from crashing, in the above-mentioned multiple conversion relationships, a maximum value is determined as the final The conversion relationship, such as 0.9976 in Figure 4.

Then, the superimposed forecasted TPS multiplied by the conversion relationship K can be obtained in the above, and the credible forecasted TPS of the business system in each forecast period can be obtained. For example, the credible predicted TPS of the business system every month in the next year, or the credible predicted TPS of every day in the next year can be obtained. FIG. 5 shows a schematic diagram of a possible trusted prediction TPS for each prediction period.

Divide the trusted predicted TPS of each forecast period by the stand-alone TPS to determine the server capacity of the business system in each forecast period.

Specifically, the stand-alone TPS can be determined as follows:

method one

According to the server performance index collected by the business system in each historical period, the stand-alone TPS of the server is determined. The server performance index is an important index to measure the stability of the service provided by a business system. Including the server's CPU usage, memory usage, IO usage, and disk usage, etc., the collection time period is the whole year, the time unit is second level, and the collection method is the system daemon process collection method. You can set a certain threshold condition to judge the stand-alone TPS of the server. For example, when the CPU utilization rate reaches 80%, the TPS at the corresponding moment is taken as the TPS that can be carried by a single server, that is, the stand-alone TPS.

way two

With the help of the pressure test system, the capacity pressure test of the business system server can also be used to obtain the stand-alone TPS of the server under the condition that the transaction success rate and delay are normal.

General business systems have a log cold backup function, and the generated log files can be uploaded to the cloud for storage. The disk usage rate of the local server is not very high, so the performance of the disk can be optimized through code and business logic. within consideration.

So far, the server capacity of the business system in each forecast period has been obtained. Figure 6 shows a possible schematic diagram of the server capacity of the business system in each month of the next year. The operation and maintenance personnel can obtain the results of the server capacity based on To prepare for procurement, you can also perform one-click dynamic expansion and contraction on a monthly basis, saving resources and operation and maintenance costs on the premise of ensuring that the performance of the system server can carry all business needs.

Optionally, after obtaining the credible forecast TPS of each forecast period, the credible forecast TPS can also be compared with the business development demand, and the promotion rhythm can be adjusted in time according to the comparison result. For example, if the credible forecast TPS in January is 40,000, but the company's business development demand for the business system is that the highest TPS in January reaches 50,000, then in order to meet the business development demand, you can increase the promotion efforts by increasing the second The number of accounts is used to improve the trusted prediction TPS. If the credible forecast TPS exceeds the business promotion expectation, the promotion efforts can be reduced and the number of newly added second accounts can be reduced.

However, the above solution does not take into account various software and hardware failures that may exist in a business system, so the obtained server capacity is not enough to cope with various environmental changes that may occur in the business system. Therefore it can be further optimized.

Fig. 7 shows a schematic diagram of a possible computer room environment of the business system, including 3 computer rooms, each of which is equipped with 2 physical machines, and several servers are deployed on each physical machine. It is assumed that a maximum number of servers (n sets) are deployed on each physical machine. Server capacity can be optimized from the following three perspectives.

(1) Consider the transaction loss threshold for a single failure of a single physical machine

Generally, physical machines have a certain failure rate. To measure whether a business system service is up to standard, a single failure transaction loss threshold (B _{single time} ) is usually set, which is the maximum transaction volume that can be affected by a single physical machine failure. .

来表示，那么单物理机损失的交易量应该小于单次故障交易损失阈值(B _单次)。 Among them, T _max is the trusted predicted TPS, N is the server capacity, s is the maximum failure time of a single physical machine, and n is the number of servers deployed on each physical machine. Then in the above formula, the credible predicted TPS is divided by the server capacity of the business system, and the TPS that each server in the business system should carry can be obtained, and then multiplied by n to obtain the TPS that each single physical machine in the business system should handle TPS, because we believe that generally speaking, after s seconds of failure, the operation and maintenance personnel will find the failed physical machine and repair or replace it, then the transaction volume lost by the single physical machine within s seconds of the failure can be used

To represent, then the transaction volume lost by a single physical machine should be less than the transaction loss threshold for a single failure (B _single ).

Transform the above formula to obtain the first range of server capacity:

(2) Consider single room failure disaster recovery

Considering the situation of disaster recovery, the business system will be deployed in several different computer rooms to avoid the failure of a single computer room to cause the entire system to be paralyzed. That is to say, if any computer room fails, the remaining machines must be able to support the transaction peak. In this way, the number of machines in a computer room needs to be redundant. The more average the machines deployed in each computer room, the fewer redundant machines, so it is best is the average distribution of all computer rooms, that is

也应能够满足该业务系统的可信预测TPS的需求，即

其中，T _单机为单机TPS。 Among them, L _{computer room} is the number of computer rooms of the business system. When a computer room fails, the remaining server capacity is

It should also be able to meet the needs of the trusted predictive TPS of the business system, namely

Among them, T _stand-alone is stand-alone TPS.

Transform the above formula to obtain the second range of server capacity:

(3) Consider server performance stability

Considering the stability of server performance, the server will not be fully loaded to provide services, that is, the single machine will not be allowed to carry the amount of T _{single machine} for a long time, which will easily cause transaction timeouts, so it is necessary to set a server that can provide services stably The highest load rate ∈, the maximum load rate of the machine planning on the production of the business system of the present invention is 80%, that is to say, in any failure situation, the load rate of the server will be lower than ∈,

Transform the above formula to obtain the third range of server capacity:

Optionally, in actual use, the server capacity can be determined respectively within the first range, the second range or the third range; the server capacity can also be determined within the first range and the second range, and can also be determined within the first range The server capacity is determined within the scope and the third scope, which is not limited in this embodiment of the present invention. The following shows the formula for determining the server capacity according to the first range and the third range:

Then, after obtaining the credible predicted TPS(T _max ) of each forecast period, bring T _max into the above relational expression, a range of server capacity that should be deployed in this business system can be obtained, and the optimized TPS can be determined according to this range. server capacity.

The method for predicting server capacity provided by the embodiment of the present invention takes into account the different rules of change of user activity in different time periods, and determines the predicted activity of each forecast period by analyzing the change rules of historical activity in different historical periods , so that dynamic expansion and contraction can be performed in the process of providing services by the business system. On the basis of ensuring the optimal capacity of the business system, fewer resource pools can be equipped, and on the premise of ensuring the normal and effective operation of the business system, it can reduce Identify resource overprovisioning and waste due to system load.

Due to the division of different statistical dimensions, the user's trading habits are analyzed according to different statistical dimensions, and different laws are determined for different statistical dimensions, and further corrections are made through the conversion relationship, and the credible predicted TPS value obtained will be more accurate. Accurate and closer to real user behavior.

Taking into account various abnormal conditions such as single physical machine failure, single computer room failure, and the highest load rate of a single server, the entire business system can be quickly isolated and recovered in the event of daily failures, providing more stable services.

Based on the same technical concept, FIG. 8 exemplarily shows the structure of an apparatus for predicting server capacity provided by an embodiment of the present invention, and the structure can execute a process of predicting server capacity.

As shown in Figure 8, the device specifically includes:

A determining unit 801, configured to:

For any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period;

According to the regular characteristics of each statistical dimension, predict the predicted transaction volume per second TPS of each statistical dimension in each forecast period, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the superposition forecast of the business system in each forecast period TPS;

Convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the TPS of the business system in each historical period The relationship between the superimposed historical TPS and the credible historical TPS is obtained by analyzing the relationship; the superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; the credible historical TPS is obtained by analyzing the business system in It is obtained by analyzing the TPS of each historical period;

The server capacity of the service system in each forecast period is determined through the credible predicted TPS of the service system in each forecast period.

Since user activity will change in different historical periods, and in different statistical dimensions, the change rules are different, so by analyzing the historical activity of each historical period in different statistical dimensions in the business system, we can get The regular characteristics of the historical activity of each statistical dimension in each historical period can more accurately predict the predicted TPS of each statistical dimension in each forecast period. Then, by superimposing the predicted TPS of different statistical dimensions, the superimposed predicted TPS of the entire business system can be further determined. The superimposed predicted TPS obtained in this way can more accurately reflect the changing rules of different statistical dimensions in each forecast period. In order to further improve the accuracy of the forecast, through the analysis of the credible historical TPS and superimposed historical TPS of the business system in each historical period, the conversion relationship between the two is determined, and the credible predicted TPS is obtained according to the conversion relationship and the superimposed predicted TPS. The credible predicted TPS is more accurate. Therefore, the server capacity of the business system in each forecast period can be accurately predicted.

Optionally, the determining unit is specifically configured to:

The quotient of the historical TPS of the statistical dimension in each historical period and the number of first accounts in each historical period is determined as the historical activity of the statistical dimension in each historical period.

Optionally, the determining unit is specifically configured to:

Fitting the historical activity of the statistical dimension in each historical period to obtain a fitting curve corresponding to each preset function;

A fitting curve corresponding to each preset function conforming to a preset condition is determined as a regular feature.

Optionally, the determining unit is specifically configured to:

If it is determined that there is no fitting curve that meets the preset conditions, the historical activity of the statistical dimension in each historical period is segmented into intervals;

For any interval segment, the floating coefficient of each historical activity in the interval segment is determined as the regular feature of the interval segment; the floating coefficient is determined according to the floating ratio of each historical activity; The floating ratio is the ratio of the difference between the highest value of historical activity and the lowest value of historical activity in the interval segment to the lowest value of historical activity.

Optionally, the determining unit is specifically configured to:

Determine the forecast activity of each statistical dimension in each forecast period through the regular characteristics of each statistical dimension;

The product of the forecast activity of each statistical dimension in each forecast period and the second account number in the same forecast period is determined as the predicted TPS of each statistical dimension in each forecast period.

Optionally, the determining unit is specifically configured to:

Collect the server performance indicators of the business system in each historical period, and determine the TPS carried by the server when the server performance indicators meet the set conditions as the stand-alone TPS of the server;

The server capacity of the service system in each forecast period is determined through the trusted forecast TPS and the stand-alone TPS.

Optionally, the determining unit is specifically configured to:

The first range of server capacity is determined according to the following formula:

Wherein, T _max is the trusted predicted TPS, N is the server capacity, s is the maximum failure time of a single physical machine, n is the number of servers deployed on a single physical machine, and B is the transaction loss threshold _once ;

The server capacity is determined within the first range of server capacities.

Optionally, the determining unit is specifically configured to:

The second range of server capacity is determined according to the following formula:

Wherein, the L _{computer room} is the number of computer rooms of the business system, and the T _stand-alone is the single-machine TPS;

The server capacity is determined within the first range of server capacities and the second range of server capacities.

Optionally, the determining unit is specifically configured to:

The third range of server capacity is determined according to the following formula:

Wherein, ∈ is the highest load rate;

The server capacity is determined within the first range of server capacities and the third range of server capacities.

Based on the same technical concept, the embodiment of the present application provides a computer device, as shown in FIG. 9 , including at least one processor 901, and a memory 902 connected to at least one processor. The processor is not limited in the embodiment of the present application. As for the specific connection medium between 901 and memory 902, the bus connection between processor 901 and memory 902 in FIG. 9 is taken as an example. The bus can be divided into address bus, data bus, control bus and so on.

In this embodiment of the present application, the memory 902 stores instructions executable by at least one processor 901, and at least one processor 901 executes the instructions stored in the memory 902 to perform the steps of the above method for predicting server capacity.

Among them, the processor 901 is the control center of the computer equipment, which can use various interfaces and lines to connect various parts of the computer equipment, by running or executing the instructions stored in the memory 902 and calling the data stored in the memory 902, so as to make predictions server capacity. Optionally, the processor 901 may include one or more processing units, and the processor 901 may integrate an application processor and a modem processor. The tuner processor mainly handles wireless communication. It can be understood that the foregoing modem processor may not be integrated into the processor 901 . In some embodiments, the processor 901 and the memory 902 can be implemented on the same chip, and in some embodiments, they can also be implemented on independent chips.

The processor 901 can be a general processor, such as a central processing unit (CPU), a digital signal processor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array or other programmable logic devices, discrete gates or transistors Logic devices and discrete hardware components can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the methods disclosed in connection with the embodiments of the present application may be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

The memory 902, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs and modules. The memory 902 may include at least one type of storage medium, for example, may include a flash memory, a hard disk, a multimedia card, a card memory, a random access memory (Random Access Memory, RAM), a static random access memory (Static Random Access Memory, SRAM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Magnetic Memory, Disk , CD, etc. Memory 902 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 902 in the embodiment of the present application may also be a circuit or any other device capable of implementing a storage function, and is used for storing program instructions and/or data.

Based on the same technical concept, an embodiment of the present invention also provides a computer-readable storage medium, the computer-readable storage medium stores a computer-executable program, and the computer-executable program is used to make the computer execute the prediction server listed in any of the above-mentioned methods capacity method.

Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

Obviously, those skilled in the art can make various changes and modifications to the application without departing from the spirit and scope of the application. In this way, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims (12)

A method for predicting server capacity, comprising: For any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period; According to the regular characteristics of each statistical dimension, predict the predicted transaction volume per second TPS of each statistical dimension in each forecast period, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the superposition forecast of the business system in each forecast period TPS; Convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the TPS of the business system in each historical period The relationship between the superimposed historical TPS and the credible historical TPS is obtained by analyzing the relationship; the superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; the credible historical TPS is obtained by analyzing the business system in It is obtained by analyzing the TPS of each historical period; The server capacity of the service system in each forecast period is determined through the credible predicted TPS of the service system in each forecast period. The method according to claim 1, wherein the historical activity of any statistical dimension in each historical period is obtained by the following methods, including: The quotient of the historical TPS of the statistical dimension in each historical period and the number of first accounts in each historical period is determined as the historical activity of the statistical dimension in each historical period. The method according to claim 1, wherein determining the regular characteristics of the historical activity of the statistical dimension in each historical period comprises: Fitting the historical activity of the statistical dimension in each historical period to obtain a fitting curve corresponding to each preset function; A fitting curve corresponding to each preset function conforming to a preset condition is determined as a regular feature. The method according to claim 1, wherein determining the regular characteristics of the historical activity of the statistical dimension in each historical period comprises: If it is determined that there is no fitting curve that meets the preset conditions, the historical activity of the statistical dimension in each historical period is segmented into intervals; For any interval segment, the floating coefficient of each historical activity in the interval segment is determined as the regular feature of the interval segment; the floating coefficient is determined according to the floating ratio of each historical activity; The floating ratio is the ratio of the difference between the highest value of historical activity and the lowest value of historical activity in the interval segment to the lowest value of historical activity. The method according to claim 1, characterized in that, predicting the predicted TPS of each statistical dimension in each prediction period through the regular characteristics of each statistical dimension, comprising: Determine the forecast activity of each statistical dimension in each forecast period through the regular characteristics of each statistical dimension; The product of the forecast activity of each statistical dimension in each forecast period and the second account number in the same forecast period is determined as the predicted TPS of each statistical dimension in each forecast period. The method according to any one of claims 1-5, characterized in that determining the server capacity of the service system in each forecast period is determined through the trusted predicted TPS of the service system in each forecast period, including: Collect the server performance indicators of the business system in each historical period, and determine the TPS carried by the server when the server performance indicators meet the set conditions as the stand-alone TPS of the server; The server capacity of the service system in each forecast period is determined through the trusted forecast TPS and the stand-alone TPS. The method according to claim 6, wherein determining the server capacity of the service system in each forecast period through the trusted predicted TPS and the stand-alone TPS includes: The first range of server capacity is determined according to the following formula:

Wherein, T _max is the trusted predicted TPS, N is the server capacity, s is the maximum failure time of a single physical machine, n is the number of servers deployed on a single physical machine, and B is the transaction loss threshold _once ; The server capacity is determined within the first range of server capacities. The method according to claim 7, characterized in that determining the server capacity of the service system in each prediction period through the trusted predicted TPS and the stand-alone TPS includes: The second range of server capacity is determined according to the following formula:

Wherein, the L _{computer room} is the number of computer rooms of the business system, and the T _stand-alone is the single-machine TPS; The server capacity is determined within the first range of server capacities and the second range of server capacities. The method according to claim 7, characterized in that determining the server capacity of the service system in each prediction period through the trusted predicted TPS and the stand-alone TPS includes: The third range of server capacity is determined according to the following formula:

Wherein, ∈ is the highest load rate; The server capacity is determined within the first range of server capacities and the third range of server capacities. A device for predicting server capacity, characterized in that it includes: Identify units for: For any statistical dimension in the business system, analyze the historical activity of the statistical dimension in each historical period, and determine the regular characteristics of the historical activity of the statistical dimension in each historical period; According to the regular characteristics of each statistical dimension, predict the predicted transaction volume per second TPS of each statistical dimension in each forecast period, and superimpose the predicted TPS of the same forecast period in each statistical dimension to obtain the superposition forecast of the business system in each forecast period TPS; Convert the superimposed predicted TPS of the business system in each forecast period into the credible predicted TPS of the business system in each forecast period according to the conversion relationship; wherein, the conversion relationship is the TPS of the business system in each historical period The relationship between the superimposed historical TPS and the credible historical TPS is obtained by analyzing the relationship; the superimposed historical TPS is obtained by superimposing the historical TPS of the same historical period in each statistical dimension; the credible historical TPS is obtained by analyzing the business system in It is obtained by analyzing the TPS of each historical period; The server capacity of the service system in each forecast period is determined through the credible predicted TPS of the service system in each forecast period. A computing device, comprising: memory for storing computer programs; A processor, configured to call the computer program stored in the memory, and execute the method according to any one of claims 1 to 9 according to the obtained program. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer-executable program, and the computer-executable program is used to make a computer execute the method according to any one of claims 1-9.

Images