JP7331407B2 - Container activation host selection device, container activation host selection system, container activation host selection method and program - Google Patents

Description

The present invention relates to a container activation host selection device, a container activation host selection system, a container activation host selection method, and a program.

Virtual machines and containers are currently used in many computer systems. A computer on which a container runs is called a container host, and multiple containers may run on one container host. For example, if containers running high-load applications are concentrated on a certain container host, the performance of the applications running in those containers will be affected. Therefore, when starting a container, it is necessary to appropriately select the container host and balance the load among the container hosts. Techniques for selecting a container host in consideration of CPU (Central Processing Unit) and memory loads have already been provided. For example, in Patent Document 1, resource requirements for CPU and memory are described in a definition document describing containers to be started, and container hosts that satisfy the requirements are selected according to a probabilistic distribution scheme or a round-robin system. A method for selecting is disclosed. Kubernetes, which is a container orchestration tool, selects a container host for starting a container by filtering and prioritizing. By using this tool, it is possible to start containers so that the load of CPU and memory is evenly distributed among container hosts.

In addition, in Patent Literature 2, existing virtual machines are classified into types according to their configuration by items such as CPU, memory, network, storage, OS, and middleware, and history information indicating the load status is accumulated for each type. Then, when deploying a new virtual machine, the degree of similarity between the new virtual machine and the existing virtual machine is calculated from the degree of matching of the above items, and the degree of similarity and the accumulated load history information of the existing machine are used to create the new virtual machine. A method of estimating the load of each and determining a destination physical machine based on the result of the prediction is disclosed.

JP 2016-057898 A JP 2017-538204 A

The method described in Patent Document 1 targets only CPUs and memories for load balancing. For this reason, the block IO load and the network IO load vary among container hosts. In addition, the method described in Patent Document 2 does not take into consideration the degree of contribution of each item to various loads. Therefore, for example, even if the load of a new virtual machine is predicted based on the load history of an existing virtual machine of the same type that matches each item of CPU, memory, network, storage, OS, and middleware, the new virtual machine If an application running on a virtual machine differs from that of the same type of virtual machine, there can be a large difference in block IO and network IO load. Therefore, even if the method of selecting a physical machine for arranging a virtual machine described in Patent Document 2 is applied to the selection of a container host, variations in block IO load and network IO load among container hosts cannot be eliminated.
There is a need for a method of selecting a container host on which to launch a container so that various loads are evenly distributed among the container hosts.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a container activation host selection device, a container activation host selection system, a container activation host selection method, and a program that solve the above-described problems.

According to one aspect of the present invention, a container activation host selection device includes predicted values of a plurality of types of loads due to applications running in containers running on each of a plurality of container hosts, and activation containers, which are containers to be activated. predicting the total of the plurality of types of loads when the startup container is started for each of the container hosts, based on the predicted values of the plurality of types of loads caused by the applications running in the a container host control unit that selects the smallest container host and decides to start the startup container on the selected container host ; identification information of an image that is the source of creation of the startup container; identification information of the image; a table for storing information in association with each other to identify an application running in the startup container; an application information determination unit that identifies the running application; and the application running in the boot container and the running container based on the actual values of the plurality of types of load caused by the applications running in the container running on the container host. and a load information prediction unit that predicts a plurality of types of the load due to the applications running in the startup container, and the application information determination unit identifies the application running in the startup container when there is a startup request for the startup container. Also, the application running in the container running in each of the plurality of container hosts is specified, and the load information prediction unit specifies the plurality of types of loads of the application running in the specified startup container and the specified load. The container host controller predicts the plurality of types of loads of the applications running in the containers running on each of the plurality of container hosts, and the container host control unit predicts all the loads running on the container hosts for each of the plurality of container hosts. totaling the predicted values of the plurality of types of loads of the applications running on the containers of the above for each of the plurality of types; is added, and the container host having the smallest sum of the loads for each of the plurality of types after the addition is determined to start the start-up container.

Further, according to another aspect of the present invention, a container activation host selection system obtains a plurality of types of loads by the above-described container activation host selection device and an application running in a container running on its own machine. and one or a plurality of container hosts provided with a container load information collection unit that notifies the container activation host selection device.

According to another aspect of the present invention, there is provided a container activation host selection method executed by a computer, which predicts a plurality of types of loads by applications running in containers running on each of a plurality of container hosts. and the predicted values of the plurality of types of load caused by the application running in the startup container, which is the container to be started, the plurality of types of load when the startup container is started for each of the container hosts. predicting and selecting the container host having the smallest sum of the plurality of types of loads, and determining to start the boot container on the selected container host; and identification information of an image from which the boot container is created; , and a table for storing image identification information and application information in association with each other, identifying an application running in the startup container, identifying the image identification information from which the container is created, and the table. a step of identifying an application running in the container based on , and an application running in the boot container based on the actual values of the plurality of types of the loads of the applications running in the container running in the container host; and a step of predicting a plurality of types of the load by applications running in the running container, and when there is a start request for the start container, the step of specifying the application running in the start container. in the step of identifying applications running in containers running on each of the plurality of container hosts, and performing prediction, the plurality of types of applications running in the specified startup containers; and the plurality of types of loads of applications running in containers running on each of the identified plurality of container hosts; totaling the predicted values of the plurality of types of loads of the applications running in all the containers running on the host for each of the plurality of types; The container activation host selection method includes adding the loads for each of the plurality of types and determining to activate the activation container on the container host having the smallest sum of the loads for the plurality of types after addition.

Further, according to another aspect of the present invention, a program stores, in a computer, predicted values of a plurality of types of load due to applications running in containers running on each of a plurality of container hosts, and a container to be started. Predicting the plurality of types of loads when the startup container is started for each of the container hosts based on the predicted values of the plurality of types of loads due to the applications running in a certain startup container, and predicting the plurality of types of loads selecting the container host having the smallest total of , and determining to start the boot container on the selected container host; identifying the image from which the boot container is created; a table for storing information in association with each other to identify an application running in the startup container; a step of identifying an application to be run; and an application to be run in the startup container and the running container based on the actual values of the plurality of types of load by the application running in the container running in the container host. and a step of predicting a plurality of types of the load by an application, and when there is an activation request to the activation container, in the step of identifying the application, the application running in the activation container is identified, and the In the step of identifying an application running in a container running on each of a plurality of container hosts and performing the prediction, the plurality of types of loads of the application running in the specified startup container and the specified in the step of predicting and determining the plurality of types of loads of applications running in containers running on each of the plurality of container hosts, all containers running on the container hosts for each of the plurality of container hosts; totaling the predicted values of the plurality of types of loads of the applications running in the plurality of types for each of the plurality of types; Then, the container host having the smallest sum of the loads for each of the plurality of types after addition determines to start the start-up container.

According to the present invention, loads on multiple container hosts can be evenly distributed.

1 is a first block diagram illustrating an example of a container activation host selection system according to one embodiment of the present invention; FIG. 2 is a second block diagram illustrating an example of a container activation host selection system according to one embodiment of the present invention; FIG. 4 is a first flow chart showing an example of processing according to an embodiment of the present invention; FIG. 5 is a second flow chart showing an example of processing according to an embodiment of the present invention; FIG. It is a figure which shows an example of the application information accumulation table which shows an example of the process by one Embodiment of this invention. 3 is a third flow chart showing an example of processing according to an embodiment of the present invention; It is a figure which shows an example of the application estimation information accumulation table which shows an example of the process by one Embodiment of this invention. FIG. 11 is a fourth flow chart showing an example of processing according to an embodiment of the present invention; FIG. FIG. 11 is a fifth flowchart showing an example of processing according to one embodiment of the present invention; FIG. It is a figure which shows an example of the load information accumulation table which shows an example of the process by one Embodiment of this invention. FIG. 11 is a sixth flow chart showing an example of processing according to an embodiment of the present invention; FIG. FIG. 11 is a seventh flow chart showing an example of processing according to an embodiment of the present invention; FIG. FIG. 11 is an eighth flow chart showing an example of processing according to an embodiment of the present invention; FIG. 1 is a diagram showing the minimum configuration of a container activation host selection system according to one embodiment of the present invention; FIG. 1 is a diagram showing an example of a hardware configuration of a container activation host selection system according to an embodiment of the present invention; FIG.

A container activation host selection system according to one embodiment will be described below with reference to FIGS. 1 to 15. FIG.
(Description of configuration)
1 and 2 are first and second block diagrams respectively showing an example of a container activation host selection system according to one embodiment of the present invention.
As shown in FIG. 1 , the container activation host selection system 30 includes a client 001 , a container activation host selection device 100 and one or more container hosts 200 . The client 001 has a container activation request unit 010 . The container activation host selection device 100 includes a container activation request reception unit 101, a container host control unit 102, an application information determination unit 110, an application information storage unit 111, an application information estimation unit 112, and an application information reception unit 113. , a load information prediction unit 120 , a load information storage unit 121 , and a load information reception unit 122 . The container host 200 includes a container control unit 210 , an application information collection unit 211 , a container load information collection unit 212 and one or more containers 220 .

FIG. 2 shows a configuration example when there are two container hosts 200 . The following description is based on this configuration example. As shown in FIG. 2, the container activation host selection system 30 includes a client 001, a container activation host selection device 100, and container hosts 200a and 200b. For convenience of explanation, it is assumed that one container 220a is running on the container host 200a and one container 220b is running on the container host 200b. The functional units provided in the container host 200a and the container host 200b are the same as those described with reference to FIG. Each functional unit has the following functions.

(Client 001)
The container activation request unit 010 notifies the container activation request reception unit 101 of the container activation host selection device 100 of the container activation request of the client 001 .

(Container activation host selection device 100)
The container activation request reception unit 101 receives a container activation request from the container activation request unit 010 . The container activation request reception unit 101 notifies the container host control unit 102 of the received container activation request.

The container host control unit 102 notifies the application information determination unit 110 of the container information, requests information of an application operating in the container (hereinafter referred to as application information), and acquires the application information. The container host control unit 102 notifies the load information prediction unit 120 of the application information and acquires a load prediction value. The container host control unit 102 also acquires identification information of the containers 220a and 220b operating on the respective container hosts 200a and 200b from the container control units 210a and 210b, and further acquires application information operating on the containers 220a and 220b. The container host control unit 102 notifies the load information prediction unit 120 of the application information and acquires the load information. Then, the container host control unit 102 selects a container host to start the container based on the load prediction values of the containers 220a and 220b currently in operation and the load prediction value of the container for which the start request has been made. 210a or 210b is notified of a container activation request.

The application information determining unit 110 determines which application is running in the container based on the application information for each container accumulated in the application information storage unit 111 . Further, when the application information does not exist in the application information storage unit 111, the application information determination unit 110 notifies the application information estimation unit 112 of the container information, and obtains a result of estimation as to what kind of application is running. The application information determination unit 110 notifies the container host control unit 102 of application information indicating applications that operate in the container.

The application information storage unit 111 stores, for each container, information about applications that operate in the container.

When the application that runs in the container is unknown (when the application information does not exist in the application information storage unit 111), the application information estimation unit 112 determines the application that runs in the container from the operation history information of the image that is the source of the container creation. make a guess.

The application information reception unit 113 acquires application information running in the running containers 220 a and 220 b and stores the information in the application information storage unit 111 .

The load information prediction unit 120 predicts the application load based on the information accumulated in the load information storage unit 121 .
The load information storage unit 121 stores information used for application load prediction.

The load information reception unit 122 accumulates load information. The load information receiving unit 122 acquires the load information of the containers 220a and 220b from the container load information collecting units 212a and 212b, respectively, and acquires the application information corresponding to the containers from the application information determining unit 110. Stored in the information storage unit 121 .

(Container hosts 200a and 200b)
The container control unit 210a receives an activation instruction from the container host control unit 102 and activates the container 220a. In addition, the container control unit 210a notifies the container host control unit 102 of information on the container 220a currently in operation. The same applies to the container control section 210b.
The application information collection unit 211a notifies the application information reception unit 113 of application information operating in the container 220a. The same applies to the application information collection unit 211b.
The container load information collection unit 212a periodically collects load information of containers operating in the container host 200a and notifies the load information reception unit 122 of the load information. The same applies to the container load information collection unit 212b.

(Description of operation)
The operation of the container activation host selection system 30 will be described in detail below.
FIG. 3 is a first flowchart illustrating an example of processing according to an embodiment of the invention.
FIG. 3 shows an outline of the operation of the container activation host selection system 30. As shown in FIG.
First, the container activation request unit 010 notifies the container activation request to the container activation request reception unit 101 (step S101). The container activation request specifies an identification ID of an image from which a container is created (hereinafter referred to as an image ID).
Next, the container host control unit 102 notifies the application information determination unit 110 of the image ID, and acquires the application information from the application information determination unit 110 (step S102).

Next, the container host control unit 102 notifies the load information prediction unit 120 of application information and the like, and acquires a load prediction value from the load information prediction unit 120 (step S103).
Next, the container host control unit 102 requests the load prediction values of the container hosts 200a and 200b and acquires the respective load prediction values (step S104).

Next, the container host control unit 102 selects a container host to start the container (step S105). For example, when the load of the container host 200a is smaller than that of the container host 200b, the container host control unit 102 selects the container host 200a.
Next, the container host control unit 102 notifies the container control unit 210a of the selected container host 200a of a start request for the container 220a, and starts the container 220a (step S106). The same applies when the container host 200b is selected.

Next, the application information collection unit 211 notifies the application information reception unit 113 of information on the application that operates in the started container 220a. The application information reception unit 113 updates the application information corresponding to the image ID of the container 220a stored in the application information storage unit 111 with the acquired information (step S107).

Next, the container load information collection units 212a and 212b periodically notify the load information reception unit 122 of the load information of the containers 220a and 220b in operation. The load information reception unit 122 notifies the application information determination unit 110 of the image ID of the container, and obtains application information corresponding to the containers 220a and 220b. Then, the load information receiving unit 122 stores load information for the application in the load information storage unit 121 .

Next, the process of acquiring the application corresponding to the image ID in step S102 will be described in detail with reference to FIGS. 4 and 5. FIG.
FIG. 4 is a second flowchart illustrating an example of processing according to an embodiment of the invention.
FIG. 5 is a diagram showing an example of an application information accumulation table showing an example of processing according to an embodiment of the present invention.
First, the application information determination unit 110 queries the application information record corresponding to the image ID from the application information accumulation table 400 stored in the application information storage unit 111 (step S201). FIG. 5 illustrates an application information accumulation table 400. As shown in FIG. For example, the image ID "cef95d42a67e" is stored in association with the application "nginx" that runs in the container indicated by this image ID.

Next, the application information determination unit 110 determines whether or not there is a record of the image ID in the application information accumulation table 400 (step S202).
If the record exists (step S202; Yes), the application information determination unit 110 reads the application information corresponding to the image ID from the application information accumulation table 400 and notifies it to the requesting container host control unit 102 (step S203). ).

If the record does not exist (step S202; No), the application information determination unit 110 notifies the application information estimation unit 112 of the image ID and requests estimation of application information. The application information estimation unit 112 estimates application information (step S204). Next, the application information estimation unit 112 notifies the application information determination unit 110 of the application estimation result. The application information determination unit 110 notifies the container host control unit 102 of the estimated information of the application corresponding to the image ID (step S205).

Next, with reference to FIGS. 6 and 7, the process of estimating the application in step S204 will be described in detail.
FIG. 6 is a third flowchart illustrating an example of processing according to an embodiment of the invention.
FIG. 7 is a diagram showing an example of an application estimation information accumulation table showing an example of processing according to an embodiment of the present invention.
First, the application information estimation unit 112 inquires the application estimation information accumulation table 500 stored in the application information storage unit 111 illustrated in FIG. 7 and the record of the application estimation information corresponding to the image ID (step S301). FIG. 7 illustrates an application estimation information accumulation table 500. As shown in FIG. For example, the image ID "siselksi3l23" is stored in association with application guess information "httpd" that is estimated to run in the container indicated by this image ID.
Next, application information estimation unit 112 determines whether or not a record exists (step S302).

If the record exists (step S302; Yes), the application information estimation unit 112 reads application estimation information corresponding to the image ID from the application estimation information accumulation table 500, and notifies the application information determination unit 110 (step S303). .

If the record does not exist (step S302; No), the application information estimation unit 112 acquires the operation history of the image. The operation history can be acquired using, for example, a command provided by docker (“docker history”). The operation history includes an operation log, time information, and image ID information at the time corresponding to the operation. Note that docker is container execution platform software provided as open source (step S304). Next, the application information estimation unit 112 extracts an application installation operation from the operation history (step S305). For example, the application information estimation unit 112 extracts package installation commands (yum install, apt-get install) from the operation history, and obtains information about what software has been installed. For example, if there is an operation history of "apt-get install tomcat", it is determined that "tomcat" has been installed.
The application information estimation unit 112 determines whether there is a history of installation operations (step S306).

If there is a history (step S306; Yes), the application information estimation unit 112 associates the installation target with the image ID as application estimation information and stores it in the application estimation information accumulation table 500 (step S307).
If the history does not exist (step S306; No), the application information estimation unit 112 stores information indicating that the application is unknown in the application estimation information accumulation table 500 (step S308).
The application information estimation unit 112 notifies the application information determination unit 110 of the estimation result of the application or that it is unknown (step S309).
In this embodiment, only the installation operation history is simply extracted, but the installation target may be different from the application, such as a library. To exclude this, an application list table may be prepared in advance, and only the applications listed in the list may be extracted. Alternatively, an application corresponding to the image ID may be guessed by separately preparing a sandbox environment, starting a container from the image in an isolated state, and obtaining a list of processes started in the container.

Next, referring to FIG. 8, the update processing of the application information corresponding to the image ID in step S107 will be described in detail.
FIG. 8 is a fourth flowchart illustrating an example of processing according to an embodiment of the invention.
For example, assume that the container host 200a is selected as a container startup destination, and a new container 220a2 is started in the container host 200a. The application information collection unit 211a executes the process list acquisition command on the namespace of the container 220a2. The process list acquisition command is specifically "nsenter--target <$PID>--pid-ps". The application information collection unit 211a identifies that the process acquired from the process list is an application running in the container 220a2, and acquires the application indicated by the process as application information (step S401).
Next, the application information collection unit 211a notifies the application information reception unit 113 of the image ID of the container 220a2 and the application information acquired in step S401 (step S402).
Next, the application information reception unit 113 adds a record in which the image ID and the application information are associated with each other to the application information accumulation table 400 stored in the application information storage unit 111 (step S403).

Next, referring to FIGS. 9 and 10, the process of accumulating application load information in step S108 will be described in detail.
FIG. 9 is a fifth flowchart illustrating an example of processing according to an embodiment of the invention.
FIG. 10 is a diagram showing an example of a load information accumulation table showing an example of processing according to one embodiment of the present invention.
First, the container load information collection unit 212a periodically acquires performance information of the container 220a. For example, the container load information collection unit 212a executes a "docker status" command provided by container execution infrastructure software docker, and acquires load information of CPU, memory, block IO, and network IO for each container. When the container 220a2 is activated in the container host 200a, the container load information collection unit 212a similarly periodically acquires various load information of the container 220a2. Similarly, the container load information collecting unit 212b periodically acquires the performance information of the container 220b (step S501).
Next, the container load information collection units 212a and 212b notify the load information reception unit 122 of the acquired container information (image ID of the container 220a and the like) and the acquired load information (step S502).
Next, the load information reception unit 122 notifies the application information determination unit 110 of the image ID, and acquires application information corresponding to the image ID (step S503).
Next, the load information reception unit 122 stores load information corresponding to the application information in the load information storage unit 121 based on the information received in steps S502 and S503. Specifically, the load information storage unit 121 adds information that associates the container ID, the image ID, the application information, and the load information to the load information accumulation table 600 illustrated in FIG. 10 together with the time when the load information is acquired. (step S504).

Next, with reference to FIG. 11, the process of acquiring the application load prediction value in step S103 will be described in detail.
FIG. 11 is a sixth flowchart illustrating an example of processing according to an embodiment of the invention.
In this embodiment, if the execution source (tenant) of the application is the same, the load is predicted on the assumption that the same application is subjected to the same periodical load.
Although this embodiment targets a single-tenant environment, it can be extended to a multi-tenant environment as described later. Also, prediction is made with the periodicity of one week. For periodicity prediction, for example, the technique disclosed in Japanese Patent Laid-Open No. 2013-214171 can be used.
First, the load information prediction unit 120 acquires the load information for the prediction period of the CPU, memory, block IO, and network IO corresponding to the application from the load information storage unit 121 (step S601). For example, the load information prediction unit 120 acquires load information for one week from the load information accumulation table 600 ( FIG. 10 ) stored in the load information storage unit 121 . If there is data acquired from a plurality of containers 220 for the target application, the load information prediction unit 120 acquires all of them from the load information accumulation table 600 .
The load information prediction unit 120 performs statistical processing on the acquired load information. For example, the load information prediction unit 120 calculates an average value for each load of CPU, memory, block IO, and network IO. The load information prediction unit 120 notifies the container host control unit 102 of the average value of each load of the CPU, memory, block IO, and network IO as a load prediction value (step S602).

Next, with reference to FIG. 12, the process of acquiring the load prediction of each container host in step S104 will be described in detail.
FIG. 12 is a seventh flowchart illustrating an example of processing according to an embodiment of the invention.
First, the container host control unit 102 requests and acquires list information (list of container IDs) of containers operating in each container host from the container control units 210a and 210b (step S701). For example, in the configuration example illustrated in FIG. 2, the container control unit 210a acquires identification information (container ID) of the container 220a as list information. If multiple containers 220 are in operation, the container IDs of all of them are acquired. The same applies to the container control section 210b. Also, the container host control unit 102 acquires an image ID from the identification information of each container included in the list information, for example, by issuing a predetermined command.
Next, the container host control unit 102 notifies the application information determination unit 110 of the acquired image ID of the container, and acquires application information (step S702).
Next, the container host control unit 102 notifies the load information prediction unit 120 of the container ID, image ID, and application information for the containers currently running on each of the container hosts 200a and 200b. A load prediction value is requested and obtained (step S703). The load information prediction unit 120 acquires, for example, one week's worth of load information from the load information accumulation table 600 illustrated in FIG. The container host control unit 102, for example, uses the acquired load information (actual value) for one week as the predicted load value for one week.
Next, the container host control unit 102 sums the load prediction values acquired in S703 for each container host, and sums the prediction values of each load of CPU, memory, block IO, and network IO for all applications by load, and is calculated (step S704).

Next, referring to FIG. 13, the process of selecting a container host for starting a container in step S105 will be described in detail.
FIG. 13 is an eighth flow chart illustrating an example of processing according to an embodiment of the present invention.
First, the container host control unit 102 adds up the application load prediction value acquired in step S103 and the load prediction value of each container host acquired in step S104 for each CPU, memory, block IO, and network IO (step S801). That is, the container host control unit 102 calculates the values of various loads when additionally starting the container 220 to be started for each of the container hosts 200a and 200b.
Next, the container host control unit 102 determines whether any of the total values of the loads calculated in step S801 exceeds the threshold during a certain period of time (step S802). If the threshold is exceeded, the container host is excluded from selection. However, if the threshold is exceeded for all container hosts, the next step is executed for all container hosts.
Next, the container host control unit 102 selects a container host to start the container using the following evaluation function F (step S803). The container host control unit 102 selects a container host indicated by the evaluation function F. FIG.

Here, x in equation (1) indicates the container host (container host 200a, 200b in the configuration example of FIG. 2. y indicates the type of load (CPU, memory, block IO, network IO in this embodiment). Axy indicates the average value of predicted load values in the prediction period for container host x and load y.By indicates the average value of predicted load values in the prediction period for all container hosts and load y. 1) selects the container host 200 that has the smallest sum of average predicted load values of CPU, memory, block IO, and network IO in the prediction period.

According to this embodiment, when the container 220 is started, the application that runs on the container 220 is determined, and the CPU, memory, block IO, and network IO consumed by the application to be started and the application running on the container host 200 are calculated. , the container host 200 predicted to have a low load is selected, and the container 220 is started on that container host 200 . As a result, with respect to the load on the container host 200 that starts the container 220, the load on the CPU, memory, block IO, and network IO can be evenly distributed.

In the above embodiment, the CPU, memory, block IO, and network IO obtained by executing the command ("docker status") provided by the container execution platform software docker are used as load indicators. Instead of load or memory load, or in addition to CPU, memory, block IO, network IO, other values such as the number of threads and the number of file descriptors per container 220 may be used as indicators of load.

In the above-described embodiment, when selecting the container host 200, the CPU, memory, block IO, and network IO loads are equally weighted, but the weight of each load may be adjusted.

Further, for example, for the CPU load and memory load, the condition for selecting a container host is that these values are equal to or less than a predetermined threshold value, while the block IO load and network IO load are predicted by the method of the present embodiment. to select the container host with the lowest expected load.

In the above embodiment, a single tenant environment is targeted, but an ID that identifies a tenant (hereinafter referred to as a tenant ID) is further associated with application information that distinguishes load information and managed. By doing so, it is also possible to target a multi-tenant environment. In that case, the tenant ID is added to the column items of the container activation request notification, the load information notification, and the load information accumulation table 600, and the load information corresponding to the tenant ID and application information is recorded and used for prediction.

FIG. 14 is a diagram showing the minimum configuration of a container activation host selection device according to one embodiment of the present invention.
As shown in FIG. 14, the container activation host selection device 100 comprises at least a container host controller 102 .
For each of a plurality of container hosts, the container host control unit 102 predicts a plurality of types of load based on the application running in the container running on each container host, and the application running in the startup container that is the container to be started. Based on the predicted values of multiple types of loads, predict multiple types of loads when launching a startup container on each of multiple container hosts, and decide to launch the startup container on the container host with the lowest load among multiple types. do. Multiple types of load may include block IO (disk IO) or network IO.

FIG. 15 is a diagram showing an example of the hardware configuration of a container activation host selection system according to one embodiment of the present invention.
A computer 900 includes a CPU 901 , a main memory device 902 , an auxiliary memory device 903 , an input/output interface 904 and a communication interface 905 . The container activation host selection device 100 and the container host 200 described above are implemented in the computer 900 . The operation of each functional unit described above is stored in the auxiliary storage device 903 in the form of a program. The CPU 901 reads out the program from the auxiliary storage device 903, develops it in the main storage device 902, and executes the above processing according to the program. Also, the CPU 901 secures a storage area in the main storage device 902 according to the program. In addition, the CPU 901 secures a storage area for storing data being processed in the auxiliary storage device 903 according to the program.

It should be noted that, in at least one embodiment, secondary storage device 903 is an example of non-transitory tangible media. Other examples of non-transitory tangible media include magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, etc. that are connected via the input/output interface 904 . Moreover, when this program is delivered to the computer 900 via a communication line, the computer 900 receiving the delivery may develop the program in the main storage device 902 and execute the above process. Also, the program may be for realizing part of the functions described above. Furthermore, the program may be a so-called difference file (difference program) that implements the above-described functions in combination with another program already stored in the auxiliary storage device 903 .

In addition, it is possible to appropriately replace the components in the above-described embodiments with well-known components without departing from the scope of the present invention. Moreover, the technical scope of the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the present invention.

30 Container activation host selection system 001 Client 010 Container activation request unit 100 Container activation host selection device 101 Container activation request reception unit 102 Container host control unit 110 Application information determination unit 111 Application information storage unit 112 Application information estimation unit 113 Application information reception unit 120 Load information prediction unit 121 Load information storage unit 122 Load information reception units 200, 200a, 200b Container hosts 210, 210a, 210b Container control units 211, 211a, 211b Application information collection units 212, 212a, 212b Container load information collection units 220, 220a, 220b Container 900 Computer 901 CPU
902 Main storage device 903 Auxiliary storage device 904 Input/output interface 905 Communication interface

Claims (10)

Predicted values of multiple types of loads by applications running in containers running on each of a plurality of container hosts and predicted values of multiple types of loads by applications running in startup containers that are containers to be started based on the container host, predict a plurality of types of the load when the startup container is started for each of the container hosts, select the container host with the smallest total of the plurality of types of the load, and a container host control unit that decides to start the startup container ;
The application running in the boot container is specified based on the identification information of the image from which the boot container is created and the table storing the identification information of the image and the information of the application in association with each other. an application information determination unit that identifies an application running in the container based on the identification information of the image to be created and the table;
Prediction of the plurality of types of load by the application running in the startup container and the application running in the running container based on the actual values of the plurality of types of load by the application running in the container running in the container host a load information prediction unit that performs
with
When there is an activation request for the activation container,
The application information determination unit identifies an application running in the startup container and identifies an application running in a container running in each of the plurality of container hosts,
The load information prediction unit predicts the plurality of types of loads of the applications running on the identified start-up containers and the plurality of types of applications running on the containers running on each of the plurality of identified container hosts. predict the load and
The container host control unit sums the predicted values of the plurality of types of loads of the applications running in all the containers running on the container hosts for each of the plurality of container hosts, and further sums the predicted values for each of the plurality of types. by adding the loads of the plurality of types of the applications running in the startup container for each of the plurality of container hosts to , and starting the startup container on the container host with the smallest sum of the loads of the plurality of types after addition Determining container launch host selection device. The multiple types of loads include disk IO loads.
2. The container activation host selection device according to claim 1. The plurality of types of loads include network IO loads.
3. The container activation host selection device according to claim 1 or 2. The container host control unit compares one or more of the added loads of the plurality of types with a predetermined threshold value, and if there is a load exceeding the threshold value, the container host is , Exclude the starting container from being started,
The container activation host selection device according to any one of claims 1 to 3. a load information receiving unit that acquires from the container host a plurality of actual load values of the applications running in the containers running in the container host, and stores the actual load values for each of the applications in a storage unit;
The container activation host selection device according to any one of claims 1 to 4, further comprising: The container host control unit calculates an average value of all the container hosts for each of the plurality of types of loads after addition, and calculates the same average value from each of the plurality of types of loads after addition for each container host. subtracting the average value associated with the type of load, dividing the subtracted value by the average value to calculate a first evaluation value, and calculating a weighted sum of the first evaluation values for all of the plurality of types of loads; determining to launch the launch container on the container host with the lowest
The container activation host selection device according to any one of claims 1 to 5. If the table does not store the application information about the application running in the boot container , the application running in the boot container is estimated based on the installation history of the application for the image from which the boot container is created. application information guessing unit,
further comprising
The load information prediction unit predicts the estimated plurality of types of loads of the application.
The container activation host selection device according to any one of claims 1 to 6 . a container activation host selection device according to any one of claims 1 to 6;
one or a plurality of container hosts comprising a container load information collection unit that acquires a plurality of types of the loads by applications running in the containers running on the own device and notifies the container activation host selection device;
A container launch host selection system including. A computer-implemented container launch host selection method comprising:
Predicted values of multiple types of loads by applications running in containers running on each of a plurality of container hosts and predicted values of multiple types of loads by applications running in startup containers that are containers to be started based on the container host, predict a plurality of types of the load when the startup container is started for each of the container hosts, select the container host with the smallest total of the plurality of types of the load, and determining to launch the launch container ;
The application running in the boot container is specified based on the identification information of the image from which the boot container is created and the table storing the identification information of the image and the information of the application in association with each other. identifying an application running in the container based on the identification information of the image to be created and the table;
Prediction of the plurality of types of load by the application running in the startup container and the application running in the running container based on the actual values of the multiple types of load by the application running in the container running in the container host. and
When there is an activation request for the activation container,
In the step of identifying the application, identifying the application running in the startup container and identifying the application running in the container running in each of the plurality of container hosts;
In the step of predicting, the plurality of types of loads of applications running on the identified startup containers and the plurality of types of applications running on containers running on each of the plurality of identified container hosts. and predict the load of
In the determining step, for each of the plurality of container hosts, the predicted values of the plurality of types of loads of the applications running in all the containers running on the container hosts are totaled for each of the plurality of types, and the totaled value is obtained. by adding the loads of the plurality of types of the applications running in the startup container for each of the plurality of container hosts to , and starting the startup container on the container host with the smallest sum of the loads of the plurality of types after addition Determines how the container launch host is selected. to the computer,
Predicted values of multiple types of loads by applications running in containers running on each of a plurality of container hosts and predicted values of multiple types of loads by applications running in startup containers that are containers to be started based on the container host, predict a plurality of types of the load when the startup container is started for each of the container hosts, select the container host with the smallest total of the plurality of types of the load, and determining to launch the launch container ;
The application running in the boot container is specified based on the identification information of the image from which the boot container is created and the table storing the identification information of the image and the information of the application in association with each other. identifying an application running in the container based on the identification information of the image to be created and the table;
Prediction of the plurality of types of load by the application running in the startup container and the application running in the running container based on the actual values of the multiple types of load by the application running in the container running in the container host. and
When there is an activation request for the activation container,
In the step of identifying the application, identifying the application running in the startup container and identifying the application running in the container running in each of the plurality of container hosts;
In the step of predicting, the plurality of types of loads of applications running on the identified startup containers and the plurality of types of applications running on containers running on each of the plurality of identified container hosts. and predict the load of
In the determining step, for each of the plurality of container hosts, the predicted values of the plurality of types of loads of the applications running in all the containers running on the container hosts are totaled for each of the plurality of types, and the totaled value is obtained. by adding the loads of the plurality of types of the applications running in the startup container for each of the plurality of container hosts to , and starting the startup container on the container host with the smallest sum of the loads of the plurality of types after addition A program that executes a decision process .

Images