CN103778474A - Resource load capacity prediction method, analysis prediction system and service operation monitoring system - Google Patents

Abstract

The invention provides a resource load prediction method, an analysis and prediction system, and a business operation monitoring system. Wherein the method includes: respectively obtaining the historical data sequence of business elements and the historical data sequence of resource load; generating the expected business element according to the historical data sequence of business elements; , establishing a correlation function between the business element and the resource load; and generating the expected resource load corresponding to the expected business element according to the correlation function. The embodiment of the present invention fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, and accurately predicts the resource load required by different applications in the future according to the load balancing characteristics of different applications. The resource deployment and maintenance mechanism of the cloud computing platform is improved, which helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

Description

Resource load forecasting method, analysis and forecasting system, and business operation monitoring system

technical field

The present invention relates to communication technology and network technology, in particular to a method for predicting resource load, an analysis and prediction system, and a business operation monitoring system.

Background technique

With the maturity of cloud technology, the cloud computing industry tends to analyze and predict the load of basic resources (such as the CPU utilization rate of virtual machines and the amount of memory resources), and the cloud computing system architecture shown in Figure 1 appears. As shown in Figure 1, the cloud computing system architecture includes: a service management system, a basic resource management system, and an analysis and prediction system. Among them, the service management system mainly implements functions such as SLA (Service-Level Agreement, Service Level Agreement) management, service catalog management, billing management, and customer management. The basic resource management system mainly realizes functions such as resource monitoring, resource control, resource dynamic allocation, and resource dynamic migration. The analysis and prediction system mainly realizes the functions of data analysis, trend analysis, forecast and alarm. The data analysis function is mainly to preprocess the collected basic resource load; the trend analysis function is mainly aimed at the analyzed basic resource load, and predicts the change trend of the basic resource load at a certain time in the future according to a specific prediction model and prediction algorithm ; The prediction and alarm function is mainly aimed at the abnormal state resource data of trend prediction, which is sent to the service management system through network communication, and the service management system notifies each user in the form of SMS or email.

In the prior art, the analysis and prediction system only uses the resource load monitored by the basic resource management system as historical data, and predicts the future resource load without considering other aspects such as the load characteristics of each application in the cloud computing platform. The impact of external factors on the resource load, the prediction accuracy is low.

Contents of the invention

Aspects of the present invention provide a resource load forecasting method, an analysis and forecasting system, and a business operation monitoring system, so as to improve the accuracy of forecasting.

A first aspect of the present invention provides a resource load prediction method, including:

Respectively obtain the historical data sequence of business elements and the historical data sequence of resource load, wherein the monitoring time carried by each business element information in the historical data sequence of business elements is respectively the same as that carried by the information of each resource load in the historical data sequence of resource load The monitoring time is the same, the business element information is the information that the business operation monitoring system stores the monitored business elements and monitoring time together in the database, and the resource load information is the resource load and the resource load that the resource management system will monitor. The information stored in the database together with the monitoring time;

generating expected business elements according to the historical data sequence of the business elements;

Establishing a correlation function between business elements and resource loads according to the historical data sequence of the business elements and the historical data sequence of the resource load;

According to the association function, an expected resource load corresponding to the expected business element is generated.

The method for predicting resource load as described above, wherein, before acquiring the historical data sequence of business elements and the historical data sequence of resource load respectively, further includes:

Obtain business element information;

Judging whether the monitoring time carried by the business element information is within the preset singular day period, if so, storing the business element information into the business element singular day historical data sequence, otherwise, storing the business element information into the business element Normal daily historical data series;

Obtain resource load information;

Judging whether the monitoring time carried by the resource load information is within the preset singular day period, if so, storing the resource load information into the resource load singular daily historical data sequence, otherwise, storing the resource load information Store the normal daily historical data sequence of resource load;

Correspondingly, in the respectively acquiring the historical data sequence of the business element and the historical data sequence of the resource load, the historical data sequence of the business element is the singular day historical data sequence of the business element, and the historical data sequence of the resource load is the singular daily data sequence of the resource load. The daily historical data sequence, or the historical data sequence of the business element is a normal daily historical data sequence of the business element, and the historical data sequence of the resource load is a normal daily historical data sequence of the resource load.

The method for predicting resource load as described above, wherein, after acquiring the historical data sequence of business elements and the historical data sequence of resource load respectively, further includes:

Perform data preprocessing on the historical data sequence of the business element and the historical data sequence of the resource load respectively.

The method for predicting resource load as described above, wherein, after calculating the expected resource load corresponding to the expected service element according to the correlation function, it further includes:

storing the expected resource load in a database, so that the service management system reads the expected resource load from the database, and generates a resource deployment instruction according to the expected resource load, the resource management system The system executes resource deployment according to the resource deployment instruction.

The method for predicting resource load as described above, wherein the generating expected business elements according to the historical data sequence of the business elements includes:

Obtaining prediction errors of at least two prediction models according to the historical data sequence of the business elements;

According to the historical data sequence of the business elements, the expected business elements are calculated by using the prediction model with the smallest prediction error.

The method for predicting resource loads as described above, wherein said establishment of a correlation function between business elements and resource loads according to the historical data sequence of business elements and the historical data sequence of resource loads includes:

According to the historical data sequence of the business element and the historical data sequence of the resource load, at least two function models are used to perform function fitting processing of the business element and the resource load, and at least two candidate correlation functions are generated;

According to the historical data sequence of the business element and the historical data sequence of the resource load, calculate the fitting residual using each candidate correlation function according to the preset least square calculation algorithm;

The fitting residuals of the candidate correlation functions are compared, and the candidate correlation function with the smallest fitting residual is used as the correlation function between the business element and the resource load.

A second aspect of the present invention provides a resource load prediction method, including:

monitoring business elements;

Store the monitored business elements and monitoring time in the database, so that the analysis and prediction system can obtain the historical data series of business elements from the database, generate expected business elements according to the historical data series of business elements, and generate expected business elements according to the The historical data sequence of the business element and the historical data sequence of the resource load, establishing a correlation function between the business element and the resource load, and generating the expected resource load corresponding to the expected business element according to the correlation function;

Wherein, the monitoring time carried by each business element information in the business element historical data sequence is the same as the monitoring time carried by each resource load information in the resource load historical data sequence, and the resource load historical data sequence is determined by the The analysis and prediction system obtains from the database, and the resource load information is information that the resource management system stores in the database together with the monitored resource load and monitoring time.

According to the method for predicting resource load as described above, before the monitoring of business elements, it also includes:

Receive monitoring instruction information sent by the service management system, the monitoring instruction information carries business elements to be monitored; correspondingly

The monitoring business elements are specifically:

The business element is monitored according to the monitoring instruction information.

A third aspect of the present invention provides an analysis and prediction system, including:

The first acquisition module is used to obtain the historical data sequence of business elements and the historical data sequence of resource load respectively, wherein the monitoring time carried by each business element information in the historical data sequence of business elements is respectively the same as the historical data sequence of resource load The monitoring time carried by each resource load information is the same. The business element information is the information that the business operation monitoring system stores the monitored business elements and monitoring time together in the database. The resource load information is the resource management system. Information stored in the database together with the monitored resource load and monitoring time;

A first processing module, configured to generate expected business elements according to the historical data sequence of the business elements;

Establishing a module for establishing a correlation function between business elements and resource loads according to the historical data sequence of the business elements and the historical data sequence of the resource load;

The second processing module is configured to generate an expected resource load corresponding to the expected business element according to the association function.

The analysis and prediction system described above also includes:

The second obtaining module is used to obtain business element information;

A judging module, configured to judge whether the monitoring time carried by the business element information is within the preset singular day period, if so, generate first judgment information, otherwise, generate second judgment information;

The execution module is used to store the business element information into the business element singular day history data sequence according to the first judgment information; according to the second judgment information, store the business element information into the business element normal day history data sequence;

The second obtaining module is also used to obtain resource load information;

The judging module is also used to judge whether the monitoring time carried by the resource load information is within the preset singular day period, if so, generate third judging information, otherwise, generate fourth judging information;

The execution module is further configured to store the resource load information in the resource load singular daily historical data sequence according to the third judgment information; and store the resource load information in the resource load data sequence according to the second judgment information. Enter the normal daily historical data sequence of resource load;

Correspondingly, the historical data sequence of business elements acquired by the first acquisition unit is a historical data sequence of singular days of business elements, and the historical data sequence of resource loads is a historical data sequence of singular days of resource loads, or, the first The historical data sequence of business elements acquired by an acquisition unit is a normal day historical data sequence of business elements, and the historical data sequence of resource load is a normal daily historical data sequence of resource load.

The analysis and prediction system described above also includes:

A data preprocessing module, configured to perform data preprocessing on the historical data of the business element and the historical data of the resource load respectively.

The analysis and prediction system described above also includes:

A storage module, configured to store the expected resource load in a database, so that the service management system reads the expected resource load from the database, and generates a resource deployment instruction according to the expected resource load , the resource management system executes resource deployment according to the resource deployment instruction.

The analysis and prediction system as described above, the first processing module includes:

A first processing unit, configured to obtain forecast errors of at least two forecast models according to the historical data sequence of the business elements;

The second processing unit is configured to calculate an expected business element by using the prediction model with the smallest prediction error according to the historical data sequence of the business element.

As described above in the analysis and prediction system, the building module includes:

A fitting unit, configured to use at least two function models to perform function fitting processing of the business elements and resource loads according to the historical data sequence of the business elements and the historical data sequence of the resource load, and generate at least two candidate correlation functions ;

The third processing unit is used to calculate the fitting residual using each candidate correlation function according to the preset least square calculation algorithm according to the historical data sequence of the business element and the historical data sequence of the resource load;

The comparison unit is configured to compare the fitting residuals of each candidate correlation function, and use the candidate correlation function with the smallest fitting residual as the correlation function between the business element and the resource load.

A fourth aspect of the present invention provides a business operation monitoring system, including:

A monitoring module for monitoring business elements;

A storage module, configured to store the monitored business elements and monitoring time in a database, so that the analysis and forecasting system can obtain the historical data sequence of business elements from the database, and generate expected business elements according to the historical data sequence of business elements An element, establishing a correlation function between a business element and a resource load according to the historical data sequence of the business element and the historical data sequence of the resource load, and generating an expected resource load corresponding to the expected business element according to the correlation function;

Wherein, the monitoring time carried by each business element information in the business element historical data sequence is the same as the monitoring time carried by each resource load information in the resource load historical data sequence, and the resource load historical data sequence is determined by the The analysis and prediction system obtains from the database, and the resource load information is information that the resource management system stores in the database together with the monitored resource load and monitoring time.

The above-mentioned business operation monitoring system also includes:

The receiving module is used to receive the monitoring instruction information sent by the service management system, the monitoring instruction information carries the business elements to be monitored, and correspondingly

The monitoring module is specifically configured to monitor business elements according to the monitoring instruction information.

A fifth aspect of the present invention provides a cloud computing system architecture, including: a service management system, a database, a resource management system, the above-mentioned analysis and prediction system, and the above-mentioned business operation monitoring system.

It can be seen from the above technical solution that the embodiment of the present invention realizes accurate prediction of the future state of the resource load of the application by introducing business elements as influencing factors in the prediction method and by fitting the correlation function between the business elements and the resource load . The embodiment of the present invention fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, and accurately predicts the resource load required by different applications in the future according to the load balancing characteristics of different applications. The resource deployment and maintenance mechanism of the cloud computing platform is improved, which helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained according to these drawings without any creative effort.

FIG. 1 is a schematic structural diagram of a cloud computing system architecture in the prior art;

FIG. 2 is a flow chart of Embodiment 1 of the load prediction method provided by the present invention;

FIG. 3 is a flow chart of Embodiment 2 of the load prediction method provided by the present invention;

FIG. 4 is a flow chart of Embodiment 3 of the load prediction method provided by the present invention;

Fig. 5 is a curve schematic diagram of a specific example of the historical data sequence of business elements of GPS application;

Fig. 6 is a schematic diagram of a curve of a specific example of the historical data sequence of the resource load of the GPS application;

Fig. 7 is a schematic diagram of the expected business elements of GPS applications from January 1, 2012 to July 1, 2012 predicted by the load forecasting method provided by the embodiment of the present invention based on the historical data sequence of business elements shown in Fig. 5 ;

Figure 8 is a schematic diagram of the principle of calculating the fitting residual;

Fig. 9 shows the GPS applications from January 1, 2012 to July 1, 2012 obtained by using the load forecasting method provided by the embodiment of the present invention based on the expected business elements of the GPS application shown in Fig. A graphical representation of the expected resource load for ;

FIG. 10 is a schematic structural diagram of Embodiment 1 of the analysis and prediction system provided by the present invention;

Fig. 11 is a schematic structural diagram of Embodiment 2 of the analysis and prediction system provided by the present invention;

Fig. 12 is a schematic structural diagram of a specific example of the first processing module in the embodiment of the analysis and prediction system provided by the present invention;

Fig. 13 is a schematic structural diagram of a specific example of the building module in the embodiment of the analysis and prediction system provided by the present invention;

FIG. 14 is a schematic structural diagram of Embodiment 1 of the business operation monitoring system provided by the present invention;

FIG. 15 is a schematic structural diagram of Embodiment 1 of the cloud computing system architecture provided by the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Fig. 2 is a flow chart of Embodiment 1 of the resource load prediction method provided by the present invention. As shown in Fig. 2, the method includes:

Step 101. Acquire the historical data sequence of business elements and the historical data sequence of resource load respectively, wherein the monitoring time carried by each business element information in the historical data sequence of business elements is respectively the same as that of each resource load in the historical data sequence of resource load. The monitoring time carried by the resource load information is the same, the business element information is the information that the business operation monitoring system stores the monitored business elements and monitoring time together in the database, and the resource load information is the resource that the resource management system will monitor The load and the monitoring time are stored together in the information in the database.

Specifically, the analysis and prediction system obtains the historical data sequence of business elements and the historical data sequence of resource load respectively. Wherein, each business element information in the business element historical data sequence should correspond to each resource load information in the resource load historical data sequence. That is, when the business operation monitoring system detects a business element, the resource management system should simultaneously monitor the resource load corresponding to the business element at this time. Therefore, when the business operation monitoring system monitors the business element, it needs to record the monitoring time at the same time, and store the business element and the monitoring time together in the data. When the resource management system monitors the resource load, it also needs to record the monitoring time, and store the resource load and the monitoring time together in data. Specifically, in order to ensure the correspondence between business elements and resource loads in terms of monitoring time, the business operation monitoring system monitors business elements according to a preset time or preset cycle, and stores the monitored business elements and monitoring time together in the database middle. Likewise, the resource management system monitors the resource load at the same preset time or period as the business operation monitoring system, and stores the monitored resource load and monitoring time in the database. Wherein, the business element may specifically be: the number of registered users, the number of simultaneous online users, the number of simultaneous online requests, and the like. For example, GPS service applications can be divided into three types, which are Class A GPS service applications, Class B GPS service applications, and Class C GPS service applications. Among them, Class A GPS service applications perform relatively static load balancing in advance based on user ranges, such as number segments, and increase the division granularity of number segments when overloaded. Therefore, Class A GPS service applications mainly perform load balancing based on the number of registered users. Class B GPS service applications allocate resources relatively dynamically according to user requests. However, due to the correlation between service requests of this type of applications, a single user's request will be bound to a specific application instance within a certain period of time. Therefore, Class B GPS service applications Load balancing and resource deployment are mainly based on the number of concurrent online users. Class C GPS service applications are dynamically allocated based on user requests. Since there is no correlation between the requests of this type of applications, any request can be routed to any resource node. Therefore, Class C GPS service applications mainly perform dynamic resource application based on the number of simultaneous online requests. . The resource load may include one or more of the following: CPU usage, memory usage, hard disk usage, and network bandwidth usage.

Step 102, generating expected business elements according to the historical data sequence of the business elements.

Specifically, in analyzing the forecasting system, first, according to the historical data sequence of the business elements, obtain the forecast errors of at least two preset forecast models; then, according to the historical data series of the business elements, adopt the least predictive error Forecasting models calculate expected business elements. Wherein, according to the historical data sequence of the business elements, the prediction errors of at least two preset prediction models are obtained, which may be specifically implemented by the following method. The method includes: dividing the business element information in the business element historical data sequence according to the monitoring information carried by the business element, for example, the business element historical data sequence includes the business element information from August 1st to September 5th. According to the time information carried by the business element information, divide the business element information in the sequence into two groups, for example, the business elements from August 1 to August 31 are the first group, and the business elements from September 1 to September 5 The business element of is the second array. Then, the first array is used as historical data, the second array is used as forecast data, and the forecast model is used to calculate the forecast data corresponding to the monitoring time of each business element in the second array according to the first array. Finally, compare the predicted data with the business elements corresponding to the monitoring time in the second array to obtain the difference, and calculate the prediction error of the prediction model according to the difference. Similarly, using the above method, the prediction errors of other prediction models can be obtained.

Step 103, according to the historical data sequence of the business element and the historical data sequence of the resource load, establish a correlation function between the business element and the resource load.

Specifically, the analysis and prediction system first uses at least two function models to perform function fitting processing of the business elements and resource loads according to the historical data sequence of the business elements and the historical data sequence of the resource load, and generates at least two candidate correlation function; then, according to the historical data sequence of the business element and the historical data sequence of the resource load, calculate the fitting residual of each candidate correlation function according to the preset least square calculation algorithm; finally, compare the fitting residuals of each candidate correlation function For the fitting residual, the candidate correlation function with the smallest fitting residual is used as the correlation function between the business element and the resource load.

Step 104: Generate an expected resource load corresponding to the expected business element according to the correlation function.

Specifically, the analysis and forecasting system takes the expected business element as a calculation parameter, and then generates the expected resource load corresponding to the expected business element according to the correlation function.

The embodiment of the present invention realizes accurate prediction of the future state of the resource load of the application by introducing business elements as influencing factors in the prediction method, and by fitting the correlation function between the business elements and the resource load. This embodiment fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, accurately predicts the resource load required by different applications in the future, and improves the resource deployment of the cloud computing platform. And the maintenance mechanism helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

Furthermore, in practical applications, each application may experience a sudden increase or decrease in business elements in different time periods, for example, GPS applications, during holidays, its business elements, namely the number of registered users, the number of concurrent online users or The number of simultaneous online requests will be significantly higher than that of working days. Therefore, in order to further improve the prediction accuracy of the resource load prediction method described in the embodiment of the present invention, the resource load prediction needs to be divided into resource load prediction during odd days and resource load prediction during normal days. Among them, a strange day refers to a date on which business elements and resource loads suddenly increase or decrease several times on this day, that is, a period of time when the business elements and resource loads are different from normal days. For example, the strange days of GPS applications are legal holidays. A normal day is a date other than an odd day, for example, a normal day of a GPS application is a working day except legal holidays. Specifically, the present invention provides the second embodiment of the resource load prediction method. As shown in FIG. 3 , in addition to the above steps 101-104, the second embodiment also includes before the step 101:

Step 201, acquire business element information.

Specifically, the analysis and prediction system acquires business element information from a database, wherein the business element information includes at least a business element and monitoring time.

Step 202: Determine whether the monitoring time carried by the business element information is within the preset singular day period, if so, store the business element information in the business element singular day historical data sequence, otherwise, store the business element information Enter the normal daily historical data series of business elements.

Specifically, the analysis and prediction system judges whether the monitoring time carried by the business element information is within the preset singular day period, and if so, the analysis and prediction system stores the business element information into the business element singular day historical data sequence, otherwise , the analysis and prediction system stores the business element information into a normal day historical data sequence of the business element.

Step 203, acquiring resource load information.

Specifically, the analysis and prediction system acquires resource load information from a database, wherein the resource load information includes at least business elements and monitoring time.

Step 204: Determine whether the monitoring time carried by the resource load information is within the preset singular day period, if so, store the resource load information into the resource load singular day historical data sequence, otherwise, save the resource The load information is stored in the normal daily historical data sequence of the resource load.

Specifically, the analysis and prediction system judges whether the monitoring time carried by the resource load information is within the preset singular day period, and if so, the analysis and prediction system stores the resource load information in the resource load information singular daily historical data sequence, otherwise, the analysis and prediction system stores the resource load information into a normal daily historical data sequence of resource load information.

Correspondingly, in step 101 of the above-mentioned first embodiment, the historical data sequence of the business element and the historical data sequence of the resource load respectively acquired are specifically: the historical data sequence of the business element is the historical data sequence of the singular day of the business element, and the The historical data sequence of the resource load is the historical data sequence of the unusual day of the resource load, or the historical data sequence of the business element is the historical data sequence of the normal day of the business element, and the historical data sequence of the resource load is the historical data sequence of the normal day of the resource load .

Using steps 102 to 104 of the first embodiment above, the expected resource load during the odd day and the expected resource load during the normal day can be predicted respectively.

Still further, in the method for predicting resource load in each of the above embodiments, after the step 101, it further includes: performing data preprocessing on the historical data sequence of the business element and the historical data sequence of the resource load respectively.

Specifically, the analysis and prediction system performs data preprocessing on the acquired historical data series of business elements, and the analysis and forecasting system performs data preprocessing on the acquired historical data series of resource loads. Among them, the purpose of data preprocessing is: when the business operation monitoring system monitors the business elements, there may be problems such as monitoring omissions and abnormality of the monitored business elements. In order to avoid the influence of these abnormal data on subsequent steps, the prediction accuracy of the resource load prediction method described in this embodiment is further improved. Therefore, the analysis and prediction system needs to perform data preprocessing on the historical data sequence of the business elements and the historical data sequence of the resource load respectively, which may include the following processing contents, for example, supplementing missing values in the historical data sequence, smoothing business Noise data in the element historical data sequence and the resource load historical data sequence, identify or delete outlier data in the business element historical data sequence and the resource load historical data sequence, and the like. Among them, to supplement the missing value in the historical data sequence, the average value of the two adjacent data can be calculated according to the values of the two adjacent data before and after, and the missing value can also be supplemented by interpolation calculation.

Furthermore, in each embodiment of the resource load forecasting method described above, after the step 104, it also includes: storing the expected resource load in a database, so that the service management system reads the The resource load is expected, and a resource deployment instruction is generated according to the expected resource load, and the resource management system executes resource deployment according to the resource deployment instruction.

Fig. 4 is a flow chart of Embodiment 3 of the load prediction method provided by the present invention. As shown in Fig. 5, the method includes:

Step 301, monitor business elements.

Specifically, the business operation monitoring system monitors business elements according to a preset time or a preset cycle. Wherein, wherein, the business element may specifically be: the number of registered users, the number of concurrently online users, the number of concurrently online requests, and the like. For example, GPS service applications can be divided into three types, which are Class A GPS service applications, Class B GPS service applications, and Class C GPS service applications. Among them, Class A GPS service applications perform relatively static load balancing in advance based on user ranges, such as number segments, and increase the division granularity of number segments when overloaded. Therefore, Class A GPS service applications mainly perform load balancing based on the number of registered users. Class B GPS service applications allocate resources relatively dynamically according to user requests. However, due to the correlation between service requests of this type of applications, a single user's request will be bound to a specific application instance within a certain period of time. Therefore, Class B GPS service applications Load balancing and resource deployment are mainly based on the number of concurrent online users. Class C GPS service applications are dynamically allocated based on user requests. Since there is no correlation between the requests of this type of applications, any request can be routed to any resource node. Therefore, Class C GPS service applications mainly perform dynamic resource application based on the number of simultaneous online requests. .

Step 302, storing the monitored business elements and monitoring time in the database, so that the analysis and forecasting system can obtain the historical data sequence of the business elements from the database, and generate the expected business elements according to the historical data sequence of the business elements, According to the historical data sequence of the business element and the historical data sequence of the resource load, establish a correlation function between the business element and the resource load, and generate the expected resource load corresponding to the expected business element according to the correlation function; wherein, the The monitoring time carried by each business element information in the business element historical data sequence is the same as the monitoring time carried by each resource load information in the resource load historical data sequence, and the resource load historical data sequence is determined by the analysis and prediction system Acquired from the database, the resource load information is information that the resource management system stores in the database together with the monitored resource load and monitoring time.

Specifically, when the business operation monitoring system detects a business element, the resource management system should simultaneously monitor the resource load corresponding to the business element at this time. Therefore, when the business operation monitoring system monitors the business element, it needs to record the monitoring time at the same time, and store the business element and the monitoring time together in the data. When the resource management system monitors the resource load, it also needs to record the monitoring time, and store the resource load and the monitoring time together in data. Specifically, in order to ensure the correspondence between business elements and resource loads in terms of monitoring time, the business operation monitoring system monitors business elements according to a preset time or preset cycle, and stores the monitored business elements and monitoring time together in the database middle. Likewise, the resource management system monitors the resource load at the same preset time or period as the business operation monitoring system, and stores the monitored resource load and monitoring time in the database.

Further, in the third embodiment of the resource load prediction method above, before the step 301, it also includes: receiving the monitoring instruction information sent by the service management system, the monitoring instruction information carrying the business elements to be monitored, so that the business operation The monitoring system monitors the business elements according to the monitoring instruction information. Specifically, when the service management system receives the application deployment application sent by the user terminal, the service management system will select the corresponding business elements according to the load balancing characteristics of the application, such as the number of registered users, the number of simultaneous online users or the number of simultaneous online requests, And send monitoring instruction information to the business operation monitoring system, wherein the monitoring instruction information carries business element information. After receiving the monitoring instruction information, the business operation monitoring system starts to monitor the business elements.

The embodiment of the present invention realizes accurate prediction of the future state of the resource load of the application by introducing business elements as influencing factors in the prediction method, and by fitting the correlation function between the business elements and the resource load. This embodiment fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, accurately predicts the resource load required by different applications in the future, and improves the resource deployment of the cloud computing platform. And the maintenance mechanism helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

In order to make the resource load prediction method provided by the embodiment of the present invention more clear, the GPS application will be used as an example below.

First, third-party users (users who apply for application deployment) apply to the service management system for application deployment, that is, send a monitoring application request, and the monitoring application request carries business elements to be monitored, such as the number of concurrent online users. The service management system sends monitoring instruction information to the business operation monitoring system according to the monitoring application request, and the monitoring instruction information carries business elements to be monitored. At the same time, resource monitoring instruction information is sent to the resource management system.

Second, the business operation monitoring system monitors the number of simultaneous online users of the GPS application according to the monitoring instruction information. In this example, the business operation monitoring system stores the monitored number of simultaneous online users of the GPS application from January 2009 to January 2012 into the database. The monitoring period of the business operation monitoring system is 1 day, with a total of 1095 monitoring data, as shown in Figure 5. At the same time, the resource management system monitors the daily resource load of the GPS application from January 2009 to January 2012 according to the resource monitoring instruction information, and stores the monitored resource load in the database. The monitoring cycle of the resource management system is every day, and there are 1095 monitoring data in total, as shown in Figure 6.

Third, the analysis and prediction system predicts the resource load for the next six months based on the number of simultaneous online users and resource load of GPS applications from January 1, 2009 to January 1, 2012. From January 2009 to mid-January 2012, the information on strange days includes: statutory holidays. Except for statutory holidays, other days are normal days. Specifically, it can be seen from Figure 5 and Figure 6 that the mutation points are the number of simultaneous online users of the GPS application and the resource load of the GPS application monitored during the singular day period. For the forecast of resource load in the next half year, the specific realization is as follows:

First, the analysis and prediction system obtains the business element information monitored from January 1, 2009 to January 1, 2012 from the database, and judges whether the monitoring time carried by the 1095 business element information is within the above-mentioned singular day period, and if so , then store the business element information into the singular day historical data sequence of the business element, otherwise, store the business element information into the normal day historical data sequence of the business element. In the same way, the analysis and prediction system obtains the resource load information monitored from January 1, 2009 to January 1, 2012 from the database, and judges whether the monitoring time carried by these 1095 resource loads is within the above-mentioned singular day period, If yes, store the resource load information in the odd day historical data sequence of resource load; otherwise, store the resource load information in the normal daily historical data sequence of resource load. The analysis and forecasting system first predicts the resource load of the GPS application in the next six months of normal days according to the normal day historical data sequence of business elements and the resource load normal day historical data sequence, and then according to the business element singular day historical data sequence and the resource load singularity The daily historical data sequence predicts the resource load of GPS applications in singular days in the next six months.

Then, the analysis and forecasting system obtains the historical data sequence of business elements and the historical data sequence of resource load respectively, and generates expected business elements according to the historical data sequence of business elements.

Wherein, the historical data sequence of the business element includes a normal day historical data sequence of the business element and a singular day historical data sequence of the business element. Specifically, the normal day historical data sequence of the business element is divided into two arrays according to the monitoring time. In order to reduce the calculation complexity, this example divides the normal day historical data sequence of the business element from January 1, 2009 to December 31, 2011 The business element information of 2012 is divided into the first array, and the business element information of January 1, 2012 is divided into the second array. Using three forecasting models, namely phase space reconstruction forecasting model, gray forecasting model, and cubic exponential smoothing forecasting model, respectively based on the first array, predict the business elements on January 1, 2012, and compare the forecasted January 1, 2012 The difference between the business element on January 1 and the business element on January 1, 2012 in the second array, and calculate the prediction error of a prediction model according to the difference. The prediction error may be expressed as a percentage, for example, the percentage of the difference and the business element on January 1, 2012 in the second array. The specific calculation results are shown in Table 1.

Table 1 Comparison table of forecast errors of three forecast models

It can be seen from Table 1 that the prediction error of the phase space reconstruction prediction model is the smallest. Therefore, this example adopts the phase space reconstruction prediction model, and obtains the expected business elements according to the normal daily historical data series of the business elements. As shown in Figure 7, based on the normal day historical data series of business elements from January 1, 2009 to January 1, 2012, the expected normal day business elements from January 2, 2012 to July 1, 2012 are obtained . Similarly, based on the historical data sequence of singular days of business elements from January 1, 2009 to January 1, 2012, the expected singular day business elements from January 2, 2012 to July 1, 2012 are obtained. As shown in FIG. 7 , the expected normal day business elements and the expected odd day business elements constitute the expected business elements from January 2, 2012 to July 1, 2012.

Among them, the realization principle of the phase space reconstruction prediction model is to first determine the phase space parameters: there are two very important parameters in the phase space reconstruction process: delay time τ and embedding dimension m. Their selection is directly related to the quality of the phase space reconstruction. The C-C method is used to determine the delay time τ and the embedding dimension m by using the correlation integral, and the calculation formula is as follows:

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In the above formula (1) $h\left(r\right)=\left\{\begin{array}{c}1,r\&Greater\; Equal;0\\ 0,r<0\end{array}\right.---\left(2\right)$

It can be seen from the above formula (1) that the correlation integral is a cumulative distribution function, which represents the probability that the distance between any two points in the phase space is less than r. Here the distance between points is represented by the infinite norm of the vector difference. Define the test statistic: S(m,N,r,t)=C(m,N,r,t)-C ^m (1,N,r,t) to describe the correlation of nonlinear time series, and is given by Statistics S(m, N, r, t) to find the delay time τ and embedding dimension m.

Among them, the calculation process of the statistic S(m,N,r,t) is: decompose the time series into t non-overlapping subsequences, and t is the reconstruction time delay, namely:

x ¹ ={x(k),k=1,1+t,...,1+Nt}

x ² ={x(k),k=2,2+t,...,2+Nt}

...

x ^t ={x(k),k=t,2t,...,N} (4)

In the above formula (4), N is an integer multiple of t, then the above-defined statistic S(m, N, r, t) adopts the strategy of block average, which is the following formula:

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; t</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\frac{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}^{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\frac{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

When N →∞ in formula (5),

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; t</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}^{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

The optimal time delay t can be taken as the time point where S(m,r,t) has the smallest mutual difference between all radii r, select the largest and smallest two radii r, and define the difference:

ΔS(m,t)=max{S(m,r _j ,t)}-min{S(m,r _j ,t)} (7)

均反映了原时间序列的自相关特性，定义指标：ΔS(m,t) measures the maximum deviation of S(m,r,t) from radius r. Since ΔS(m,t) is always a positive number, the optimal time delay τ can take the time point corresponding to the first local minimum of ΔS(m,t)~t. because

Both reflect the autocorrelation characteristics of the original time series, define the index:

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; cor</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span><span\; class="notranslate">\; |</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

Find the t corresponding to the global minimum of S _cor (t) to obtain the best embedding window t _w . The embedded window method of phase space reconstruction considers that the selection of the delay time τ should not be independent of the embedding dimension m, but should depend on the embedding window t _w = (m-1)τ, from which the embedding dimension m can be calculated.

Then, phase space reconstruction prediction is performed according to the above-calculated delay time τ and embedding dimension m. According to the way of fitting attractors in phase space, it can be divided into two methods: global method and local method. The so-called global method is to use all the points in the trajectory as the fitting object, find out its law, that is, get f( ^* ), and predict the direction of the trajectory. This method is theoretically feasible, but it is difficult to make accurate predictions when the phase space trajectory is complex. The local method is to take the last point of the phase space trajectory as the center point, and use the trajectory points closest to the center point as the relevant points, and then make a fitting for these relevant points, and then estimate the direction of the next point of the trajectory, and finally from the predicted The desired predicted values are separated from the coordinates of the trajectory points of .

In local methods, phase space trajectories are predicted from the values or orientations of neighboring points to the center point. For example, if we want to predict tomorrow's weather, we can find the day that is the closest to today's natural conditions in history, and use the weather conditions of the next day of that day as tomorrow's weather forecast value. Taking the local method of first-order approximate fitting as an example, in the phase space, a small neighborhood of the nth point is used to infer the trend of the next point. The so-called first-order approximation refers to fitting the small neighborhood around the nth point with X(t+1)=a+bX(t).

In the process of phase space reconstruction, let N be the sequence length between X(t+1)=a+bX(t), and M be the number of midpoints in the phase space, then M=N-(m-1) *τ, the expression of the phase space trajectory is:

X(t+τ)=f(X(t)) (9)

Among them, X(t+τ) can be regarded as the mapping of f(X(t)), then

X(t)=[x(t),x(t+τ),...,x(t+(m-1)τ)] (10)

The mapping in the above formula (10) can be expressed as the following time series:

X(1)=[x(1),x(1+τ),...,x(1+(m-1)τ)]

X(2)=[x(2),x(2+τ),...,x(2+(m-1)τ)]

...

X(M)=[x(M),x(M+τ),...,x(N)] (11)

Let the reference vector set {X _Mi } of the central point X _M , i=1, 2,...,q, the phase point set after k steps of evolution be {X _Mi+k }, then the first-order local linear fitting as the following function:

X _Mi+k ＝a _k e+b _k X _Mi ,i=1,2,...,q (12)

According to the weighted least squares method:

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 13</span>}<span\; class="notranslate">\; )</span>$

是参考向量X_Mi的第j个元素。将上式（13）看成是关于未知数a_k和b_k的二元函数，将上述式（13）两边求偏导并化简得出：in,

is the jth element of the reference vector X _Mi. The above formula (13) is regarded as a binary function about the unknowns a _k and b _k , and the partial derivative and simplification of both sides of the above formula (13) are obtained:

$\left\{\begin{array}{c}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; Mi</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 14</span>}<span\; class="notranslate">\; )</span>$

Rewrite the above formula (14) into matrix form as:

$\left(\begin{array}{cc}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}& \mathrm{<span\; class="notranslate">\; \&beta;</span>}\\ \mathrm{<span\; class="notranslate">\; m</span>}& \mathrm{<span\; class="notranslate">\; \&alpha;</span>}\end{array}\right)\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right)<span\; class="notranslate">\; =</span>\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right)<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 15</span>}<span\; class="notranslate">\; )</span>$

in: $\mathrm{\&alpha;}=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{P}_i\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{x}_{\mathrm{Mi}}^j,$ $\mathrm{\&beta;}=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{P}_i\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{\left(x_{\mathrm{Mi}}^j\right)}^2,$ $e_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{P}_i\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{x}_{\mathrm{Mi}+k}^j{x}_{\mathrm{Mi}}^j,$ $f_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{P}_i\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{x}_{\mathrm{Mi}+k}^j$ but:

$\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right)<span\; class="notranslate">\; =</span>{\left(\begin{array}{cc}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}& \mathrm{<span\; class="notranslate">\; \&beta;</span>}\\ \mathrm{<span\; class="notranslate">\; m</span>}& \mathrm{<span\; class="notranslate">\; \&alpha;</span>}\end{array}\right)}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right)<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 16</span>}<span\; class="notranslate">\; )</span>$

According to the obtained a _k , b _k , substituting the k-step prediction formula X _M+1 = a _k e+b _k X _M , the phase point prediction value after k-step evolution can be obtained:

X _M+k ＝(x _M+k ,x _M+k+τ ，...,x _M+k+(m-1)τ ) (17)

Here, the mth element x _M+k+(m-1)τ in X _M+ k is the k-step predicted value of the original sequence.

在t＝12时取得第一个局部极小值点，因此取τ=12作为最佳时延。S_cor(t)在t＝52时取得全局最小点，即t_w=52。根据t_w=(m-1)τ，因此取m＝5，按照式(11)进行相空间重构。Based on the above phase space reconstruction prediction algorithm, after calculation

Get the first local minimum value point at t=12, so take τ=12 as the best time delay. S _cor (t) obtains the global minimum point at t=52, that is, t _w =52. According to t _w =(m-1)τ, m=5 is taken, and the phase space reconstruction is carried out according to formula (11).

The realization principle of the gray forecasting model is that the GM (1, 1) model is a dynamic sequence processing method in the gray system theory, and it is a first-order differential equation containing only a single variable.

Let x ⁽⁰⁾ be a non-negative sequence:

x ⁽⁰⁾ ={x ⁽⁰⁾ (1),x ⁽⁰⁾ (2),...,x ⁽⁰⁾ (n)}(18)

x ⁽¹⁾ is the first-order cumulative sequence of x ⁽⁰⁾ :

x ⁽¹⁾ ={x ⁽¹⁾ (1),x ⁽¹⁾ (2),...,x ⁽¹⁾ (n)} (19)

Where: x ⁽⁰⁾ (k)>0,k=1,2,..., k=1,2,...,n, then the corresponding differential equation of the GM(1,1) model is:

$\frac{{\mathrm{<span\; class="notranslate">\; dx</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; dt</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 20</span>}<span\; class="notranslate">\; )</span>$

In formula (20), α is the development gray number; u is the gray action.

利用最小二乘法求解得：

其中，parameter vector

Using the method of least squares to solve:

in,

$\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; =</span>\left|\begin{array}{cc}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right|<span\; class="notranslate">\; ,</span>$ ${\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\left|\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>\\ {\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>\\ <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\\ {\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>\end{array}\right|$

Substituting the calculated parameters α and u into formula (20) and solving it, taking x ⁽¹⁾ (0)=x ⁽⁰⁾ (1), the gray forecasting model is obtained:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; ]</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;t</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; twenty\; one</span>}<span\; class="notranslate">\; )</span>$

的模拟值，再由一次累减得到真实的预测值：The value obtained by formula (21) is the cumulative value

The simulated value of , and then the real predicted value is obtained by a cumulative subtraction:

${\hat{x}}^{\left(0\right)}(t+1)={\hat{x}}^{\left(1\right)}(t+1)-{\hat{x}}^{\left(1\right)}\left(t\right)---\left(\mathrm{twenty\; two}\right)$ Where, t=1,2,...,n.

Based on the principle of the above-mentioned gray prediction model, first use the least square method to solve the parameter vector:

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; B</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; B</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\left(\begin{array}{c}\mathrm{<span\; class="notranslate">\; 0.0347</span>}\\ \mathrm{<span\; class="notranslate">\; 52.8371</span>}\end{array}\right)$

according to $\widehat{\mathrm{\&alpha;}}=\left(\begin{array}{c}a\\ u\end{array}\right)=\left(\begin{array}{c}0.0347\\ 52.8371\end{array}\right)$ It can be seen that a=0.0347, u=52.8371. Substituting the parameters a and u into the differential equation to solve, the gray prediction response function can be obtained as:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; ]</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;t</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; 1439.8629</span>}\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.0347</span>}\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1522.6809</span>}$

Among them, the posterior difference ratio is C=32.83%, and the small error probability is P=91.64%, which shows that the prediction accuracy level is good.

The realization principle of the cubic exponential smoothing forecasting model is that the principle of exponential smoothing forecasting is to use the smoothing of historical data to eliminate the influence of random factors. Triple exponential smoothing is required when the historical data series has a curvilinear tendency. The basic principle of Cubic Exponential Smoothing Forecasting Model is to estimate the quadratic polynomial parameters after the original data is processed by Cubic Exponential Smoothing, so as to establish a forecasting model. As follows:

二次指数平滑值记为

三次指数平滑值记为

其中，平滑初始值的确定采用如下公式：Let the time series be X ₁ , X ₂ , X ₃ ,...,X _n , use S to denote the exponential smoothing value, and record the exponential smoothing value in the tth period as

The quadratic exponential smoothing value is denoted as

Triple exponential smoothing is denoted as

Among them, the smooth initial value is determined using the following formula:

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}}{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; twenty\; three</span>}<span\; class="notranslate">\; )</span>$

Then the formula for calculating the exponential smoothing value is:

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; twenty\; four</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 25</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 26</span>}<span\; class="notranslate">\; )</span>$

Among them, α∈[0,1] is the smoothing coefficient, and the smoothing coefficient α can determine its reasonable size according to the actual situation. Generally, it is selected according to the minimum mean square error, that is, to perform exponential smoothing prediction on different α values, respectively calculate the mean square error, and take the smallest mean square error. The alpha value of the variance is used as the smoothing coefficient. For the index forecast value Y _t+T with a forecast period of T days and a base number of the t-th day, the mathematical model of the triple exponential smoothing method is:

Y _t+T ＝a _t +b _t T+c _t T ² (27)

Among them: a _t , b _t , c _t are smoothing coefficients, and the calculation formula is:

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>{{\mathrm{<span\; class="notranslate">\; 3</span>}\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; 3</span>}\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; 3</span>}\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 28</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; [</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 5</span>}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 29</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; [</span>{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ]</span>$

In the forecasting process of the triple exponential smoothing method, the smoothing initial value is first determined:

${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}}{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1541</span>}<span\; class="notranslate">\; ,</span>$

Secondly, the exponential smoothing prediction is performed on different α values between [0, 1] with an interval of 0.02, and the value of α=0.72 with the minimum mean square error is taken as the smoothing coefficient. Assuming that the forecast period T=1,2,...,5 days, the base is the index forecast value Y _1095+T on the t=1095th day, which is obtained by using the exponential smoothing value calculation formula Substituting the smoothing coefficient calculation formula a ₁₀₉₅ =29.3925, b ₁₀₉₅ =-0.4814, c ₁₀₉₅ =0.0212, then the calculation formula of the triple exponential smoothing method for the number of simultaneous online users is Y _1095+T =29.3925-0.4814T+0.0212T ² .

Subsequently, the analysis and forecasting system establishes a correlation function between business elements and resource loads according to the acquired normal daily historical data series of business elements and normal daily historical data series of resource loads.

Specifically, it can be observed from the historical data in Figure 6 that as the number of simultaneous online users increases, the resource load also shows a proportional increase trend. Therefore, the analysis and prediction system can respectively use the four preset growth functions to perform function fitting processing on the resource load and the number of concurrent online users. The four growth functions include: linear function type, polynomial function type, exponential function type, and power function type.

Linear function: y=ax+b;

Polynomial function: y=a ₁ x ⁿ +a ₂ x ^n-1 +a ₃ x ^n-2 +....+a _n-1 x+a _n ;

Exponential function: y=ae ^bx ;

Power function: y=ax ^b .

Among them, the choice of n value in the polynomial function should not be too high, the higher the n value, the more complicated the fitting calculation process.

In essence, the function fitting process is to substitute the business elements and resource loads with the same monitoring time in the normal day historical data series of business elements and resource load normal day historical data series into the function respectively, and solve the above function Among them, such as the coefficients a and b in the linear function, the coefficients a1, a2, a3,...an in the polynomial function, the coefficient a and the power exponent b in the exponential function and the power function. After the coefficients or exponents in the above functions are solved, four candidate correlation functions are obtained. Subsequently, according to the normal day historical data sequence of the business element and the normal day historical data sequence of the resource load, the fitting residual using each candidate correlation function is calculated according to the preset least square calculation algorithm, so as to select the fitting residual The smallest candidate correlation function is used as the correlation function between the number of simultaneous online users and the resource load.

采用四种增长型函数进行函数拟合处理后的候选关联函数及各候选关联函数的拟合残差，如表2所示。Wherein, the principle of the least square calculation algorithm is that, according to the normal daily historical data sequence of business elements and the normal daily historical data sequence of resource load carrying business elements and resource loads with the same monitoring time, n nodes can be obtained, Such as (x ₁ , y ₁ ), (x ₂ , y ₂ ), (x ₃ , y ₃ )..., (x _n , y _n ). As shown in Figure 8, the values of y ₁ ', y _{2 ', y 3 '..., y n ' corresponding to the candidate correlation functions at the same x 1 , x 2 , x 3 , ... , x n} coordinates are respectively the same as y ₁ , The sum of the squares of the differences of y ₂ , y ₃ , ..., y _n is the fitting residual. Specifically expressed as:

The candidate correlation functions and the fitting residuals of each candidate correlation function are shown in Table 2 after using four kinds of growth functions for function fitting.

Table 2. Comparison table of candidate correlation functions and fitting residuals of each candidate correlation function

It can be seen from the above table 2 that the fitting residual error of the linear function y=0.04x-45.00 is the smallest, and the linear function y=0.04x-45.00 is the correlation function between the number of simultaneous online users and the resource load of the GPS application.

Finally, according to the linear function y=0.04x-45.00, the expected resource load y is calculated by using the predicted expected business elements as calculation parameters, that is, x in the linear function. According to the expected business elements from January 2, 2012 to July 1, 2012 described in Figure 8, calculate the expected resource load from January 2, 2012 to July 1, 2012 as shown in Figure 9 .

Those of ordinary skill in the art can understand that all or part of the steps for implementing the above method embodiments can be completed by program instructions and related hardware. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above-mentioned method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

As shown in FIG. 10 , a schematic structural diagram of Embodiment 1 of the analysis and prediction system provided by the present invention. As shown in the figure, the analysis and prediction system includes: a first acquisition module 1, a first processing module 2, a building module 3 and a first Two processing modules4. Wherein, the first acquisition module 1 is used to respectively acquire the historical data sequence of business elements and the historical data sequence of resource load, wherein, the monitoring time carried by each business element information in the historical data sequence of business elements is respectively related to the resource load The monitoring time carried by each resource load information in the volume historical data sequence is the same. The business element information is the information that the business operation monitoring system stores the monitored business elements and monitoring time together in the database. The resource load information is The resource management system stores the monitored resource load and the monitoring time together into the information in the database. The first processing module 2 is used for generating expected business elements according to the historical data sequence of the business elements. The establishment module 3 is used to establish a correlation function between business elements and resource loads according to the historical data sequence of business elements and the historical data sequence of resource loads. The second processing module 4 is configured to generate an expected resource load corresponding to the expected business element according to the association function.

The analysis and prediction system described in the embodiment of the present invention realizes accurate prediction of the future state of the resource load of the application by introducing business elements as influencing factors in the prediction process and by fitting the correlation function between the business elements and the resource load. This embodiment fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, accurately predicts the resource load required by different applications in the future, and improves the resource deployment of the cloud computing platform. And the maintenance mechanism helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

Further, as shown in FIG. 11 , the above embodiment of the analysis and prediction system further includes: a second acquisition module 5 , a judgment module 6 and an execution module 7 . Wherein, the second acquiring module 5 is used for acquiring business element information. The judging module 6 is used to judge whether the monitoring time carried in the business element information is within the preset singular day period, and if so, generate first judging information, otherwise, generate second judging information. The execution module is used to store the business element information into the business element singular day historical data sequence according to the first judgment information; according to the second judgment information, store the business element information into the normal day of the business element historical data series. The second acquiring module 7 is also used for acquiring resource load information. The judging module is also used to judge whether the monitoring time carried in the resource load information is within the preset singular day period, and if so, generate third judging information, otherwise, generate fourth judging information. The execution module is further configured to store the resource load information in the resource load singular day historical data sequence according to the third judgment information; and store the resource load information in the resource load data sequence according to the second judgment information. Normal daily historical data series of resource load. Correspondingly, the historical data sequence of business elements acquired by the first acquisition unit 1 is a historical data sequence of singular days of business elements, and the historical data sequence of resource loads is a historical data sequence of singular days of resource loads, or, the The historical data sequence of business elements acquired by the first acquisition unit is a normal day historical data sequence of business elements, and the historical data sequence of resource load is a normal daily historical data sequence of resource load.

Still further, the above-mentioned embodiment of the analysis and prediction system further includes: a data preprocessing module. Wherein, the data preprocessing module is used to perform data preprocessing on the historical data sequence of the business element and the historical data sequence of the resource load respectively.

Furthermore, the above-mentioned embodiment of the analysis and prediction system further includes: a storage module. The storage module is configured to store the expected resource load in a database, so that the service management system reads the expected resource load from the database, and generates resource deployment according to the expected resource load An instruction, the resource management system executes resource deployment according to the resource deployment instruction.

Wherein, in the above-mentioned embodiment of the analysis and prediction system, the first processing module can be realized by adopting the structure shown in FIG. 12 . Specifically, the first processing module 2 includes: a first processing unit 21 and a second processing unit 22 . Wherein, the first processing unit 21 is configured to obtain prediction errors of at least two preset prediction models according to the historical data sequence of the business elements. The second processing unit 22 is configured to calculate an expected business element by using the prediction model with the smallest prediction error according to the historical data sequence of the business element. The establishment module 3 can be implemented with the structure shown in FIG. 13 , specifically, the establishment module 3 includes: a fitting unit 31 , a third processing unit 32 and a comparison unit 33 . Wherein, the fitting unit 31 is configured to use at least two function models to perform function fitting processing of the business elements and resource loads according to the historical data sequence of the business elements and the historical data sequence of the resource load, and generate at least two a candidate correlation function. The third processing unit 32 is configured to calculate the fitting residual using each candidate correlation function according to the preset least square calculation algorithm according to the historical data sequence of the business element and the historical data sequence of the resource load. The comparison unit 33 is configured to compare the fitting residuals of the candidate correlation functions, and use the candidate correlation function with the smallest fitting residual as the correlation function between the business element and the resource load.

As shown in FIG. 14 , it is a schematic structural diagram of Embodiment 1 of the business operation monitoring system provided by the present invention. As shown in the figure, the business operation monitoring system in this embodiment includes: a monitoring module 8 and a storage module 9 . Wherein, the monitoring module 8 is used for monitoring business elements. The storage module 9 is used to store the monitored business elements and monitoring time in the database, so that the analysis and prediction system can obtain the historical data series of business elements from the database, and generate For the expected business element, establish a correlation function between the business element and the resource load according to the historical data sequence of the business element and the historical data sequence of the resource load, and generate the expected resource load corresponding to the expected business element according to the correlation function. Wherein, the monitoring time carried by each business element information in the business element historical data sequence is the same as the monitoring time carried by each resource load information in the resource load historical data sequence, and the resource load historical data sequence is determined by the The analysis and prediction system obtains from the database, and the resource load information is information that the resource management system stores in the database together with the monitored resource load and monitoring time.

The business operation monitoring system described in the embodiment of the present invention can enable the analysis and prediction system to introduce business elements as influencing factors in the forecasting process by monitoring the business elements, and realize the application by fitting the correlation function between the business elements and the resource load. accurate prediction of the future state of the resource load. This embodiment fully considers the application deployment requirements of different users and the load balancing characteristics of each application, deploys corresponding resources for different users, accurately predicts the resource load required by different applications in the future, and improves the resource deployment of the cloud computing platform. And the maintenance mechanism helps to improve the utilization rate of the infrastructure resources of the cloud computing system.

Further, the above embodiment of the business operation monitoring system further includes a receiving module. The receiving module is used to receive the monitoring instruction information sent by the service management system, the monitoring instruction information carries the service elements to be monitored, so that the business operation monitoring system can monitor the service elements according to the monitoring instruction information.

As shown in FIG. 15 , it is a schematic structural diagram of Embodiment 1 of the cloud computing system architecture provided by the present invention. As shown in FIG. 15 , the cloud computing system architecture of this embodiment includes: a service management system 11 , a database 12 , a resource management system 13 , an analysis and prediction system 14 and a business operation monitoring system 15 . Wherein, the analysis and prediction system 14 adopts the analysis and prediction system described in the above-mentioned embodiments. The business operation monitoring system 15 adopts the business operation monitoring system described in the above-mentioned embodiments. For the specific implementation principles of the analysis and prediction system and the business operation monitoring system, refer to the related content disclosed in the corresponding embodiments above, and will not be repeated here. The service management system 11 is used to implement functions such as SLA management, service catalog management, billing management, and customer management. Among them, the SLA management function can view the SLA template and related factors, such as system alarms. The service directory management function can query, open, close and delete services. The billing management function can add and delete billing policies, as well as query, renew and settle orders. Customer management function can query and modify customer related information. The resource management system 13 is used to implement functions such as resource monitoring, resource control, dynamic configuration, and dynamic migration. Among them, the resource monitoring function mainly monitors the CPU usage, memory usage, hard disk usage, and network bandwidth usage of physical machines and virtual machines in real time. The resource control function mainly performs life cycle control such as opening, closing, restarting, suspending, resuming, hibernating, and waking up virtual machines, and performs life cycle control such as opening and restarting physical machines. The dynamic configuration function is mainly for configuring the CPU or memory when the virtual machine is running. The live migration function is mainly aimed at the fast and smooth migration of a virtual machine from one host to another when the virtual machine is running.

In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or in the form of hardware plus software functional units.

The above-mentioned integrated units implemented in the form of software functional units may be stored in a computer-readable storage medium. The above-mentioned software functional units are stored in a storage medium, including several instructions to enable a computer device (which may be a personal computer, server, or network device, etc.) or a processor (processor) to execute the methods described in various embodiments of the present invention. partial steps. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM for short), random access memory (Random Access Memory, RAM for short), magnetic disk or optical disk, etc., which can store program codes. medium.

It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence. Because of the present invention, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the foregoing embodiments, the descriptions of each embodiment have their own emphases, and for parts not described in detail in a certain embodiment, reference may be made to relevant descriptions of other embodiments. Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than 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 that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (10)

1. a resource load amount Forecasting Methodology, is characterized in that, comprising:

Obtain respectively business factor historical data sequence and resource load amount historical data sequence, wherein, the monitoring time that in described business factor historical data sequence, each business factor information is carried is identical with the monitoring time that each resource load amount information in described resource load amount historical data sequence is carried respectively;

According to described business factor historical data sequence, generate expection business factor;

According to described business factor historical data sequence and resource load amount historical data sequence, set up the correlation function of business factor and resource load amount;

According to described correlation function, generate the expection resource load amount that described expection business factor is corresponding.

2. resource load amount Forecasting Methodology according to claim 1, is characterized in that, described obtain respectively business factor historical data sequence and resource load amount historical data sequence before, also comprise:

Obtain business factor information;

Judge that monitoring time that described business factor information carries is whether within the default unusual day period, if, described business factor information is deposited in to unusual day historical data sequence of business factor, otherwise, deposit described business factor information in business factor normal day historical data sequence;

Gains resources charge capacity information;

Judge that monitoring time that described resource load amount information carries is whether within the default unusual day period, if, described resource load amount information is deposited in to unusual day historical data sequence of resource load amount, otherwise, deposit described resource load amount information in resource load amount normal day historical data sequence;

Described obtaining respectively in business factor historical data sequence and resource load amount historical data sequence, described business factor historical data sequence is unusual day historical data sequence of business factor, described resource load amount historical data sequence is unusual day historical data sequence of resource load amount, or described business factor historical data sequence is the normal day historical data sequence of business factor, described resource load amount historical data sequence is the normal day historical data sequence of resource load amount.

3. resource load amount Forecasting Methodology according to claim 1 and 2, is characterized in that, described according to described correlation function, after calculating the described expection resource load amount that described expection business factor is corresponding, also comprises:

Described expection resource load amount is stored in database, so that described service management system reads described expection resource load amount from described database, and generating resource deployment instruction according to described expection resource load amount, described resource management system is carried out resource deployment according to described resource deployment instruction.

4. according to arbitrary described resource load amount Forecasting Methodology in claim 1～3, it is characterized in that, described according to described business factor historical data sequence, generate expection business factor, comprising:

According to described business factor historical data sequence, draw the predicated error of at least two kinds of forecast models;

According to described business factor historical data sequence, adopt the forecast model of described predicated error minimum to calculate expection business factor.

5. according to arbitrary described resource load amount Forecasting Methodology in claim 1～3, it is characterized in that, described according to described business factor historical data sequence and resource load amount historical data sequence, set up the correlation function of business factor and resource load amount, comprising:

According to described business factor historical data sequence and resource load amount historical data sequence, adopt at least two function models to carry out the Function Fitting processing of described business factor and resource load amount, generate at least two candidate association functions;

According to described business factor historical data sequence and resource load amount historical data sequence, calculate the matching residual error that adopts each candidate association function according to default least-squares calculation algorithm;

The matching residual error of more each candidate association function, using the candidate association function of matching residual error minimum as described business factor and the correlation function of resource load amount.

6. a resource load amount Forecasting Methodology, is characterized in that, comprising:

Monitoring business factor;

The described business factor monitoring and monitoring time are stored in database, so that analysing and predicting system obtains business factor historical data sequence from described database, according to described business factor historical data sequence, generate expection business factor, according to described business factor historical data sequence and resource load amount historical data sequence, set up the correlation function of business factor and resource load amount, according to described correlation function, generate the expection resource load amount that described expection business factor is corresponding;

Wherein, the monitoring time that in described business factor historical data sequence, each business factor information is carried is identical with the monitoring time that each resource load amount information in described resource load amount historical data sequence is carried respectively, described resource load amount historical data sequence is obtained from described database by described analysing and predicting system, and described resource load amount information is that the resource load amount monitoring and monitoring time are together stored to the information in described database by resource management system.

7. an analysing and predicting system, is characterized in that, comprising:

The first acquisition module, for obtaining respectively business factor historical data sequence and resource load amount historical data sequence, wherein, the monitoring time that in described business factor historical data sequence, each business factor information is carried is identical with the monitoring time that each resource load amount information in described resource load amount historical data sequence is carried respectively;

The first processing module, for according to described business factor historical data sequence, generates expection business factor;

Set up module, for according to described business factor historical data sequence and resource load amount historical data sequence, set up the correlation function of business factor and resource load amount;

The second processing module, for according to described correlation function, generates the expection resource load amount that described expection business factor is corresponding.

8. analysing and predicting system according to claim 7, is characterized in that, also comprises:

The second acquisition module, for obtaining business factor information;

Judge module, for judging that monitoring time that described business factor information carries whether within the default unusual day period, if so, generates the first judgement information, otherwise, generate the second judgement information;

Execution module, for according to described the first judgement information, deposits described business factor information in business factor unusual day historical data sequence; According to described the second judgement information, deposit described business factor information in business factor normal day historical data sequence;

Described the second acquisition module, also for Gains resources charge capacity information;

Described judge module, also for judging that monitoring time that described resource load amount information carries whether within the default unusual day period, if so, generates the 3rd judgement information, otherwise, generate the 4th judgement information;

Described execution module, also, for according to described the 3rd judgement information, deposits described resource load amount information in resource load amount unusual day historical data sequence; According to described the second judgement information, deposit described resource load amount information in resource load amount normal day historical data sequence;

The described business factor historical data sequence that described the first acquiring unit obtains is unusual day historical data sequence of business factor, described resource load amount historical data sequence is unusual day historical data sequence of resource load amount, or, the described business factor historical data sequence that described the first acquiring unit obtains is the normal day historical data sequence of business factor, and described resource load amount historical data sequence is the normal day historical data sequence of resource load amount.

9. according to the analysing and predicting system described in claim 7 or 8, it is characterized in that, also comprise:

Memory module, for described expection resource load amount is stored to database, so that described service management system reads described expection resource load amount from described database, and generating resource deployment instruction according to described expection resource load amount, described resource management system is carried out resource deployment according to described resource deployment instruction.

10. a service operation supervisory system, is characterized in that, comprising:

Monitoring modular, for monitoring business factor;

Memory module, for the described business factor and the monitoring time that monitor are stored to database, so that analysing and predicting system obtains business factor historical data sequence from described database, according to described business factor historical data sequence, generate expection business factor, according to described business factor historical data sequence and resource load amount historical data sequence, set up the correlation function of business factor and resource load amount, according to described correlation function, generate the expection resource load amount that described expection business factor is corresponding;

Wherein, the monitoring time that in described business factor historical data sequence, each business factor information is carried is identical with the monitoring time that each resource load amount information in described resource load amount historical data sequence is carried respectively, described resource load amount historical data sequence is obtained from described database by described analysing and predicting system, and described resource load amount information is that the resource load amount monitoring and monitoring time are together stored to the information in described database by resource management system.

Images