JP2021105772A - Prediction management system of resource usage amount, and prediction management method of resource usage amount - Google Patents

Abstract

To predict a resource usage amount in consideration of an effect of scaling of a container.SOLUTION: A prediction management system which predicts a resource usage amount of a computer for operating an application has a calculation unit which predicts the number of containers to be scaled, by using a variation value of the resource usage amount of the application which increases or decreases in accordance with a load upon one or more containers prepared on the computer, and calculates the variation value of the resource usage amount of the application for each container after prediction, based on the number of containers after prediction.SELECTED DRAWING: Figure 10

Description

The present invention relates to a container resource usage prediction management system, and specifically, to a technique for efficiently realizing performance stabilization in a container.

In recent years, container-type virtualization technology has become common in information systems. A container in container-type virtualization is an application execution unit that shares the OS (Operating System) of a physical / virtual machine and divides resources in user space. The resources here include CPU (Central Processing Unit), memory, and network resources. Containers may be scaled up or scaled out depending on the load.

In order to plan the resources of the physical / virtual machine that runs the container, it is necessary to predict the resources used by the container.

For example, a technique for predicting future free resources based on time-series data of resource usage and resource reservation information has been proposed (see Patent Document 1).

Japanese Unexamined Patent Publication No. 2016-212609

The above-mentioned Patent Document 1 predicts the free resources of a physical machine based on the time-series data of the resource usage of each virtual machine running on the physical machine and the reserved migration information.

However, when applying the same technology to containers, the impact of container scaling cannot be considered.

One aspect of the present invention is to provide a resource usage prediction management system capable of predicting resource usage in consideration of the influence of container scaling, and a resource usage prediction management method. ..

The prediction management system according to one aspect of the present invention is a prediction management system that predicts the resource usage of a computer that operates an application, and increases or decreases according to a load on one or a plurality of containers provided on the computer. The number of containers to be scaled is predicted using the fluctuation value of the resource usage of the application, and the fluctuation value of the resource usage of the application for each container after the prediction is based on the number of containers after the prediction. It is configured as a resource usage prediction management system characterized by having a calculation unit for calculating.

According to one aspect of the present invention, the resource utilization can be predicted in consideration of the influence of the scaling of the container.

It is a figure which shows the example of the network configuration including the prediction management system in this embodiment. It is a figure which shows the hardware configuration example of the prediction management system in this embodiment. It is a figure which shows the concept of the resource utilization of the container in this embodiment. It is a figure which shows the configuration example of the application basic information table in this embodiment. It is a figure which shows the configuration example of the container arrangement latest information table in this embodiment. It is a figure which shows the configuration example of the prediction information table of the total value of the resource utilization fluctuation value in this embodiment. It is a figure which shows the structural example of the scale-out prediction information table in this embodiment. It is a figure which shows the structural example of the scale-in prediction information table in this embodiment. It is a figure which shows the configuration example of the prediction information table of the average value of the resource utilization fluctuation value in this embodiment. This is a pseudo code showing an example of the processing procedure of the resource usage prediction information calculation unit in this embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. The following description and drawings are examples for explaining the present invention, and are appropriately omitted and simplified for the sake of clarification of the description. The present invention can also be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

The positions, sizes, shapes, ranges, etc. of each component shown in the drawings may not represent the actual positions, sizes, shapes, ranges, etc., in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the position, size, shape, range and the like disclosed in the drawings.

In the following description, various information may be described by expressions such as "table" and "list", but various information may be expressed by a data structure other than these. The "XX table", "XX list", etc. may be referred to as "XX information" to indicate that they do not depend on the data structure. When expressions such as "identification information", "identifier", "name", "ID", and "number" are used in explaining the identification information, these can be replaced with each other.

When there are a plurality of components having the same or similar functions, they may be described by adding different subscripts to the same reference numerals. However, when it is not necessary to distinguish between these plurality of components, the subscripts may be omitted for explanation.

Further, in the following description, a process performed by executing a program may be described, but the program is determined by being executed by a processor (for example, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit)). Since the processed processing is appropriately performed using a storage resource (for example, memory) and / or an interface device (for example, a communication port), the main body of the processing may be a processor. Similarly, the main body of processing for executing a program may be a controller, a device, a system, a computer, or a node having a processor. The main body of the processing performed by executing the program may be an arithmetic unit, and may include a dedicated circuit (for example, FPGA (Field-Programmable Gate Array) or ASIC (Application Specific Integrated Circuit)) that performs specific processing. ..

The program may be installed from the program source into a device such as a calculator. The program source may be, for example, a program distribution server or a computer-readable storage medium. When the program source is a program distribution server, the program distribution server includes a processor and a storage resource for storing the program to be distributed, and the processor of the program distribution server may distribute the program to be distributed to other computers. Further, in the following description, two or more programs may be realized as one program, or one program may be realized as two or more programs.

FIG. 1 is a network configuration diagram including a prediction management system (hereinafter referred to as a prediction management system) 100, which is a computer system that realizes prediction of resource usage of a container.

In the network configuration diagram illustrated in FIG. 1, the prediction management system 100 is connected to the physical / virtual machines 200 to 400 via the network 11. Containers 211, 221, and 222 exist on the physical / virtual machine 200. Containers 311, 321, and 322 exist on the physical / virtual machine 300. Containers 411 and 421 exist on the physical / virtual machine 400. Containers 211, 311 and 411 belong to application 510. Containers 221, 222, 321, 322, 421 belong to application 520.

FIG. 2 shows the hardware configuration of the prediction management system 100. In this embodiment, the case where the prediction management system 100 is composed of one server is illustrated, but the program that realizes this system is distributed to one or more servers, and the same function can be obtained by communicating with each other. It may be realized.

The prediction management system 100 includes a storage device 101 composed of non-volatile storage elements such as SSD (Solid State Drive) and hard disk drive, a memory 102 composed of volatile storage elements such as RAM (Random Access Memory), and a storage device 101. The program 110 held in the memory 102 is executed by reading it into the memory 102, and is connected to the CPU 103, which is a computing device that performs overall control of the device itself and various determinations, calculations, and control processes, and network 11, and communicates with other devices. A communication device 104, which is responsible for processing, is provided.

When the above-mentioned CPU 103 executes the program 110, the function of the resource usage prediction information calculation unit 111 is realized. Details of the functions to be realized will be described later.

In the storage device 101, at least a database 120 storing various information used for processing is recorded in addition to the program 110 for implementing the functions required as the prediction management system 100 of the present embodiment. Tables 121 to 126 are recorded in the database 120. Configuration examples of tables 121 to 126 will be described later.

The physical / virtual machines 200 to 400 also have a hardware configuration as a general computer device, similar to the configuration of the prediction management system described above.

FIG. 3 is a diagram showing the concept of resource usage of the container in the present embodiment. Containers 211, 311 and 411 belonging to application 510 use resources respectively. The resources to be used are divided into fixed values and variable values. The fixed value is the amount of resources used regardless of the load on the container. The variable value is the amount of resource usage that increases or decreases according to the load of the container.

Next, a configuration example of the table used by the prediction management system 100 of the present embodiment will be described.

FIG. 4 is a diagram showing a configuration example of the application basic information table 121 in this embodiment. Each record in the application basic information table 121 corresponds to one application. The field "scale-out threshold value of the average value of the resource usage fluctuation value" is the threshold value of the average value of the resource usage fluctuation value of the container that triggers the scale-out of the container belonging to the application. The field "Scale-out target value of average resource usage fluctuation value" is the average value of resource usage fluctuation value of the container, which is the target at the time of scale-out. The field "scale-in threshold value of the average value of the resource usage fluctuation value" is the threshold value of the average value of the resource usage fluctuation value of the container that triggers the scale-in of the container belonging to the application. The field "Scale-in target value of average resource usage fluctuation value" is the average value of resource usage fluctuation value of the container, which is the target at the time of scale-in. The field "fixed resource usage value" is a fixed resource usage value of the container belonging to the application.

FIG. 5 is a diagram showing a configuration example of the container layout latest information table 122 in the present embodiment. Each record in the container placement update table 122 represents the number of containers used by an application running on a physical / virtual machine.

FIG. 6 is a diagram showing a configuration example of the prediction information table 123 of the total resource usage fluctuation value in the present embodiment. Prediction information of total resource usage fluctuation value Each record in Table 123 represents a predicted value of the total value of resource usage fluctuation value at a certain time of a certain application.

FIG. 7 is a diagram showing a configuration example of the scale-out prediction information table 124 in the present embodiment. Each record in the scale-out prediction information table 124 represents the predicted time and the amount of increase in the number of containers in the application where the scale-out is predicted.

FIG. 8 is a diagram showing a configuration example of the scale-in prediction information table 125 in the present embodiment. Each record in the scale-in prediction information table 125 represents the predicted time and the amount of decrease in the number of containers in the application where the scale-in is predicted.

FIG. 9 is a diagram showing a configuration example of the prediction information table 126 of the average value of the resource usage fluctuation value in the present embodiment. Prediction information of average resource usage fluctuation value Each record in the information table 126 represents the resource usage fluctuation value per container at a certain time of an application.

Hereinafter, the resource amount prediction procedure by the prediction management system 100 in the present embodiment will be described with reference to the figure. Various operations corresponding to the procedures described below are realized by a program read by the prediction management system 100 into a memory or the like and executed. This program consists of code for performing various operations described below.

FIG. 10 is a pseudo code showing an example of the processing procedure of the resource usage prediction information calculation unit 111 in the present embodiment.

In step 1, the resource usage prediction information calculation unit 111 updates the application basic information table 121 based on the information of the container existing in the physical / virtual machine.

In step 2, the resource usage prediction information calculation unit 111 updates the container placement latest information table 122 based on the information of the containers existing in the physical / virtual machine.

In step 3, the resource usage prediction information calculation unit 111 obtains the predicted value of the resource usage by applying a predetermined prediction algorithm, and updates the prediction information table 123 of the total resource usage fluctuation value. As the above-mentioned algorithm adopted for such prediction, an algorithm corresponding to a method generally used in time series data analysis such as SARIMA (Seasonal Auto Regressive Integrated Moving Average) model and Holt-Winters method is used. good.

In steps 4 to 28, the resource usage prediction information calculation unit 111 repeats the same processing for all the applications existing in the application basic information table 121.

In step 5, the resource usage prediction information calculation unit 111 stores the value of the “scale-out threshold value of the average resource usage fluctuation value” of the application basic information table 121 related to the relevant application in the variable variable_A.

In step 6, the resource usage prediction information calculation unit 111 stores the value of the “scale-out target value of the average resource usage fluctuation value” of the application basic information table 121 related to the relevant application in the variable variable_B.

In step 7, the value of “scale-in threshold value of average value of resource usage fluctuation value” in the application basic information table 121 related to the corresponding application is stored in the variable variable_C.

In step 8, the resource usage prediction information calculation unit 111 stores the value of the “scale-in target value of the average resource usage fluctuation value” of the application basic information table 121 related to the relevant application in the variable variable_D.

In step 9, the resource usage prediction information calculation unit 111 totals the number of containers belonging to the corresponding application in the container arrangement latest information table 122 and stores them in the variable variable_E.

In steps 10 to 27, the resource usage prediction information calculation unit 111 repeats the same processing in ascending order for all the times existing in the prediction information table 123 of the total resource usage fluctuation value.

In step 11, the resource usage prediction information calculation unit 111 refers to the predicted value of the corresponding application and the corresponding time in the prediction information table 123 of the total resource usage fluctuation value and stores it in the variable variable_F.

In step 12, the resource usage prediction information calculation unit 111 calculates variable_F / variable_E and stores it in variable_G. That is, the resource usage forecast information calculation unit 111 calculates the forecast value per container at the relevant time for the relevant application.

Steps 13 to 18 are executed by the resource usage prediction information calculation unit 111 only when variable_G> = variable_A. That is, in the resource usage prediction information calculation unit 111, the predicted value per container at the relevant time for the relevant application calculated in step 12 is the “average value of the resource usage fluctuation value” in the application basic information table 121 for the relevant application. If it is equal to or greater than the “scale-out threshold” value, steps 13 to 18 are executed.

In step 14, the resource usage prediction information calculation unit 111 calculates variable_F / variable_B, rounds up the value below the minor end point of the calculation result, and stores it in variable_H. That is, the resource usage prediction information calculation unit 111 refers to the resource usage fluctuation value with respect to the “scale-out target value of the average resource usage fluctuation value” of the application basic information table 121 for the relevant application stored in step 6. Calculate the number of required containers by calculating how far the predicted values of the corresponding application and the corresponding time in the prediction information table 123 of the total value are apart. In step 15, the resource usage prediction information calculation unit 111 calculates variable_H --variable_E and stores the absolute value of the calculation result in variable_I. That is, the resource usage prediction information calculation unit 111 calculates the difference between the required number of containers and the number of containers at the relevant time for the relevant application, and determines how many containers are required (insufficient). That is, calculate the number of containers that are expected to be scaled out. In step 16, the resource usage prediction information calculation unit 111 adds a record to the scale-out prediction information table 124, which is the corresponding application and the corresponding time, and the “number of containers” is variable_I.

In step 17, the resource usage prediction information calculation unit 111 stores variable_H in variable_E. That is, the required number of containers calculated in step 14 is set in the total number of containers belonging to the corresponding application in the container arrangement latest information table 122 calculated in step 9.

Steps 19 to 24 are executed only when the resource usage prediction information calculation unit 111 is variable_G <= variable_C. That is, in the resource usage prediction information calculation unit 111, the predicted value per container at the relevant time for the relevant application calculated in step 12 is the “average value of the resource usage fluctuation value” in the application basic information table 121 for the relevant application. If the value is equal to or less than the “scale-in threshold” value, steps 19 to 24 are executed. In step 20, the resource usage prediction information calculation unit 111 calculates variable_F / variable_D, rounds up the value below the minor end point of the calculation result, and stores it in variable_J. That is, the resource usage prediction information calculation unit 111 refers to the resource usage fluctuation value with respect to the “scale-in target value of the average resource usage fluctuation value” of the application basic information table 121 for the relevant application stored in step 9. Calculate the number of required containers by calculating how far the predicted values of the corresponding application and the corresponding time in the prediction information table 123 of the total value are apart.

In step 21, the resource usage prediction information calculation unit 111 calculates variable_J --variable_E and stores the absolute value of the calculation result in variable_K. That is, the resource usage prediction information calculation unit 111 calculates the difference between the required number of containers and the number of containers at the relevant time for the relevant application, and how many containers are required (surplus). , That is, calculate the number of containers that are expected to be scaled in.

In step 22, the resource usage prediction information calculation unit 111 adds a record to the scale-in prediction information table 125, which is the corresponding application and the corresponding time, and the “number of containers” is variable_K.

In step 23, the resource usage prediction information calculation unit 111 stores variable_H in variable_E. That is, the required number of containers calculated in step 20 is set in the total number of containers belonging to the corresponding application in the container arrangement latest information table 122 calculated in step 9.

In step 25, the resource usage prediction information calculation unit 111 calculates variable_F / variable_E and stores the calculation result in variable_L. That is, the resource usage prediction information calculation unit 111 calculates the predicted value per number of new containers set in step 17 or step 23 in the corresponding application and the corresponding time.

In step 26, the resource usage prediction information calculation unit 111 adds a record of the corresponding application and the corresponding time and the “predicted value” of variable_L to the prediction information table 126 of the average value of the resource usage fluctuation value.

Thus, as shown in Step 12, Step 14, Step 15, Step 17 (or Step 12, Step 20, Step 21, Step 23), etc., a computer (eg, a physical / virtual machine) that runs an application. In the prediction management system 100 that predicts the resource usage of the above, the resource usage prediction information calculation unit 111 fluctuates the resource usage of the application that increases or decreases according to the load on one or a plurality of containers provided on the computer. Using the value (Fig. 6), the number of containers to be scaled is predicted, and based on the number of containers after the prediction, the fluctuation value of the resource usage of the application for each container after the prediction (Fig. 9). Is calculated.

In addition, the resource usage prediction information calculation unit 111 includes the fluctuation value (FIG. 6) for each time for one or more applications executed on the computer, and the latest container used by the application executed on the computer. The fluctuation value per container of the application is calculated using the number of the above (FIG. 5), and when the calculated fluctuation value per container satisfies a predetermined condition (step 13 or step 19), a predetermined value is determined. The number of containers after the prediction is calculated using the fluctuation value with respect to the target value (scale-out target value or scale-in target value) of, and is set as the latest number of containers.

By the processing described above, in the resource usage prediction management system 100 of the container including the correction by the load information, the resource usage fluctuation value per container is shown in the prediction information table 126 of the resource usage fluctuation value average value. Predicted value of is stored. Further, the scale-out prediction information table 124 stores the scale-out prediction information, and the scale-in prediction information table 125 stores the scale-in prediction information. In addition to this information, the predictive management system user plans the resources of the physical / virtual machine that runs the container based on the information stored in the application basic information table 121 and the container placement update table 122. be able to. That is, in addition to the prediction based on the time-series data of the resource usage amount, the resource usage amount of the container is calculated by using only the time-series data of the resource usage amount by considering the influence of the increase / decrease in the number of containers due to the scaling of the container. It is possible to make a more accurate prediction than when making a prediction.

11 Network
100 Prediction management system
101 storage device
102 memory
103 CPU
104 Communication equipment
110 program
111 Resource Usage Forecast Information Calculation Department
120 database
121 Application basic information table
122 Container placement latest information table
123 Prediction information table for total resource usage fluctuation values
124 Scale-out forecast information table
125 Scale-in forecast information table
126 Prediction information table of average resource usage fluctuation value
200, 300, 400 physical / virtual machine
211, 221, 222, 311, 321, 322, 411, 421 containers
510, 520 applications

Claims (4)

A predictive management system that predicts the resource usage of the computer that runs the application.
The number of containers to be scaled is predicted using the fluctuation value of the resource usage of the application that increases or decreases according to the load on one or more containers provided on the computer, and the number of containers after the prediction is calculated. Based on the calculation unit, which calculates the fluctuation value of the resource usage of the application for each container after the prediction.
A resource usage forecast management system characterized by having. The calculator uses the variable value over time for one or more applications running on the computer and the number of the latest containers used by the application running on the computer to use the application. The fluctuation value per container is calculated, and when the calculated fluctuation value per container satisfies a predetermined condition, the number of the predicted containers is calculated using the fluctuation value with respect to the predetermined target value. Set as the latest number of containers,
The resource usage prediction management system according to claim 1. It is a predictive management method performed by a predictive management system that predicts the resource usage of the computer that runs the application.
The calculation unit predicts the number of containers to be scaled by using the fluctuation value of the resource usage of the application that increases or decreases according to the load on one or more containers provided on the computer.
The calculation unit calculates the fluctuation value of the resource usage of the application for each container after the prediction based on the number of containers after the prediction.
A method for predicting and managing resource usage. The calculator uses the variable value over time for one or more applications running on the computer and the latest number of containers used by the application running on the computer to use the application. Calculate the fluctuation value per container,
When the calculated fluctuation value per container satisfies a predetermined condition, the calculation unit calculates the number of the predicted containers using the fluctuation value with respect to the predetermined target value, and calculates the number of the predicted containers of the latest container. Set as a number,
The method for predicting and managing resource usage according to claim 3.

Images