JP7234702B2 - Information processing device, container placement method, and container placement program - Google Patents

Description

The present invention relates to an information processing device, a container placement method, and a container placement program, and more particularly to an information processing device, a container placement method, and a container placement program for arranging a plurality of containers.

In recent years, with the development of the Internet of Things (IoT), services are increasingly deployed on cloud systems such as SaaS (Software as a Service) and PasS (Platform as a Service). Here, container-type virtualization is mentioned as one of the methods for realizing services with high availability. By using container-based virtualization, it is possible to flexibly increase capacity by increasing the number of containers according to the number of users. Therefore, it becomes easier to realize a service with high availability.

Here, container-type virtualization is a kind of virtualization, in which a plurality of "partitions that behave like independent servers" called containers are created in one OS (Operating System), and these are used by individual users/servers. It is what you assign to a service.

In order to flexibly reinforce resources according to the development of services, and considering fault tolerance, it is desirable that this "single OS" can also be increased logically and physically. That is, there are multiple sets of environments in which multiple containers are stored in one instance (one OS).

Patent Literature 1 discloses a technique for relocating virtual instances to multiple servers. The management server according to Patent Literature 1 collects resource usage statuses from a plurality of operating servers, determines container placement destinations, and transmits a container activation or deletion command to each operating server according to the determination result.

Patent Literature 2 discloses a technique related to a distributed container optimal placement system. In the system according to Patent Document 2, when the number of containers on a server resource is the smallest and it is determined that the resource is the most vacant, the agent sends the container running on another server resource to the server resource. Request creation.

Patent Document 3 discloses a technique for automatically monitoring and allocating resources in a logically partitioned environment. In US Pat. No. 5,200,000, the resource utilization of each partition is collected, categorized into utilization zones, and resources from partitions in low utilization zones are reallocated to partitions in high utilization zones.

JP 2017-146791 A JP 2018-156465 A JP 2006-164281 A

However, the techniques according to Patent Literatures 1 to 3 have a problem that the optimization of the arrangement of containers is insufficient. The reason for this is that the technologies disclosed in Patent Documents 1 to 3 determine reallocation based on the resource usage at the time of measurement. This is because trends do not take into account cases where resource usage is not favorable.

The present disclosure has been made to solve such problems. The purpose is to provide a placement program.

An information processing device according to a first aspect of the present disclosure includes:
a collection unit that periodically collects measured values of resource usage from a computer environment in which a plurality of containers are arranged;
a time-series analysis unit that decomposes the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing;
a prediction unit that calculates a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a placement candidate calculation unit that calculates a placement candidate for containers in the computer environment based on the predicted value;
a rearrangement unit that rearranges containers in the computer environment based on the placement candidates;
Prepare.

A container arrangement method according to a second aspect of the present disclosure includes:
the computer
Periodically collect resource usage measurements from a multi-container computing environment,
Decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing,
calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
calculating placement candidates for the container in the computer environment based on the predicted value;
Based on the placement candidates, the containers are rearranged in the computer environment.

A container placement program according to a third aspect of the present disclosure includes:
a process of periodically collecting resource usage measurements from a computer environment in which a plurality of containers are arranged;
a process of decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis;
a process of calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a process of calculating placement candidates for containers in the computer environment based on the predicted values;
a process of rearranging containers in the computer environment based on the placement candidates;
run on the computer.

According to the present disclosure, it is possible to provide an information processing device, a container placement method, and a container placement program for optimizing container placement in consideration of time-series load trend characteristics.

1 is a block diagram showing the configuration of an information processing apparatus according to a first embodiment; FIG. 4 is a flow chart showing the flow of a container arrangement method according to the first embodiment; FIG. 11 is a block diagram showing the overall configuration of a cloud system according to a second embodiment; FIG. FIG. 7 is a block diagram showing the configuration of a container placement device according to a second embodiment; 9 is a flow chart showing the flow of a container arrangement method according to the second embodiment;

Embodiments of the present disclosure are described in detail below with reference to the drawings. In each drawing, the same reference numerals are given to the same or corresponding elements, and redundant description will be omitted as necessary for clarity of description.

<Embodiment 1>
FIG. 1 is a block diagram showing the configuration of an information processing apparatus 10 according to the first embodiment. The information processing device 10 is one or more computers for managing placement, monitoring, rearrangement, etc. of a plurality of containers in a computer environment. Here, the computer environment is composed of one or more computers, partitioned into a plurality of containers, independently executing applications specified for each container, and providing services for each container. . The computer environment may be, for example, a collection of computers such as a cloud system. Here, the information processing apparatus 10 includes a collection unit 11 , a time series analysis unit 12 , a prediction unit 13 , a placement candidate calculation unit 14 and a rearrangement unit 15 .

The collection unit 11 periodically collects measured values of resource usage from the computer environment. Here, the measured value is a value obtained by measuring the amount of usage of a resource (e.g., CPU (Central Processing Unit), memory, disk, network bandwidth) in the computer environment by a sensor or the like from inside or outside the computer environment, and the measurement time corresponds to It is attached. For example, every time the collection unit 11 collects measured values, it accumulates them in a storage device (not shown). become. Note that the measurement value includes measurement data of two or more measurement items (resource types). Also, the measured value may be a value obtained by measuring the amount of resource usage for each container.

The time-series analysis unit 12 decomposes the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing. Here, a known technique can be used for time-series analysis processing. Moreover, the regular fluctuation data is a data group in which a certain regularity can be found in the fluctuation pattern along the time series of the measured values among the set of the measured values. Constant regularity includes trend variation and seasonal variation. Further, the irregular fluctuation data is a data group other than the regular fluctuation data in the set of measured values.

The prediction unit 13 calculates a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule variation data. Here, the predicted value is a set of values at one or more future times for the resource usage corresponding to the measurement items of the measurement values described above. The predicted value also includes a predicted value of the load that may occur in the future on the entire computer environment. The load includes, for example, the number of processing requests for services provided by the computer environment, the processing time of the application accompanying the processing requests, and the usage amount of resources required for processing according to the processing requests. Also, the predetermined period corresponds to at least the collection interval by the collection unit 11 . Note that the prediction unit 13 may store the calculated prediction value in the storage device in association with the time.

The placement candidate calculation unit 14 calculates placement candidates of containers in the computer environment based on the predicted values. That is, the placement candidate calculation unit 14 obtains an optimal placement pattern of containers that satisfies the resource usage indicated by the predicted value. For example, it is assumed that applications corresponding to multiple services are running on a computer environment. In this case, when the load indicated by the forecast value occurs, the resource usage amount corresponding to the number of containers allocated to each service is simulated, and the allocated number of containers for each service is arranged so that the processing for the load can be realized. Candidates can be requested. Also, in the case where the computer environment includes a plurality of instances, when the load indicated by the predicted value occurs, the resource usage is simulated according to the number of containers assigned to the instances. Then, the allocated number of containers for each instance can be obtained as a placement candidate so that the processing for the load can be realized and the number of instances can be reduced. Optimal placement candidates for multiple services and multiple instances can also be similarly determined.

The rearrangement unit 15 rearranges the containers in the computer environment based on the placement candidates. The rearrangement unit 15 instructs the computer environment to allocate the instances, containers, and services indicated by the placement candidates.

FIG. 2 is a flow chart showing the flow of the container arrangement method according to the first embodiment. First, the collection unit 11 periodically collects resource usage measurement values from a computer environment in which a plurality of containers are arranged (S11). Next, the time-series analysis unit 12 decomposes the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing (S12).

Based on the rule variation data, the prediction unit 13 then calculates a predicted value of resource usage in the computer environment from now until a predetermined period from now (S13). Subsequently, the placement candidate calculation unit 14 calculates a placement candidate of the container in the computer environment based on the predicted value (S14). After that, the rearrangement unit 15 rearranges the containers in the computer environment based on the arrangement candidates (S15).

As described above, in this embodiment, first, a predicted value of future resource usage is calculated using regular fluctuation data extracted from time-series data periodically measured in a computer environment. From the fluctuation pattern along the time series, it is possible to detect not only the resource demand at one point in time, but also the long-term fluctuation trend of the load. improves. Since highly accurate prediction values can be used to derive container placement candidates, more appropriate container placement can be realized in a computer environment. Therefore, according to the present embodiment, it is possible to provide the optimum arrangement of containers in consideration of the characteristics of the chronological load trend.

The information processing apparatus 10 includes a processor, a memory, and a storage device (not shown). Further, the storage device stores a computer program in which processing of the container arrangement method according to the present embodiment is implemented. The processor then loads the computer program from the storage device into the memory and executes the computer program. Thereby, the processor implements the functions of the collection unit 11 , the time series analysis unit 12 , the prediction unit 13 , the placement candidate calculation unit 14 and the rearrangement unit 15 .

Alternatively, the collection unit 11, the time-series analysis unit 12, the prediction unit 13, the placement candidate calculation unit 14, and the rearrangement unit 15 may each be realized by dedicated hardware. Also, part or all of each component of each device may be realized by general-purpose or dedicated circuitry, processors, etc., or combinations thereof. These may be composed of a single chip, or may be composed of multiple chips connected via a bus. A part or all of each component of each device may be implemented by a combination of the above-described circuits and the like and programs. As the processor, a CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (field-programmable gate array), or the like can be used.

Further, when part or all of each component of the information processing device 10 is realized by a plurality of information processing devices, circuits, etc., the plurality of information processing devices, circuits, etc. may be centrally arranged, They may be distributed. For example, the information processing device, circuits, and the like may be implemented as a form in which each is connected via a communication network, such as a client-server system, a cloud computing system, or the like. Also, the functions of the information processing apparatus 10 may be provided in a SaaS (Software as a Service) format.

<Embodiment 2>
Here, the problem to be solved by this embodiment will be described in detail. When it comes to arranging containers in each of a plurality of instances, if the containers are not efficiently arranged, the number of instances will increase or decrease in vain. For example, if a container whose load is concentrated during a specific time period is disproportionately allocated to a certain instance, it becomes necessary to increase the number of instances during that time period. Securing such unnecessary resources also affects the cost of the service operation side.

Here, in kubernetes, which is an orchestration tool for container-type virtual environments, there is a setting that specifies the instance (node) where the container is placed. However, in services using IoT, it is conceivable that the load is concentrated in a certain time period or a certain season for each target industry. Therefore, an operation such as setting each time is required.

In addition, the seasonal fluctuations are expected to change due to the increase in users and the growth of services (for example, the addition of target sensors), so it is necessary to arrange based on the latest trends. However, analyzing trends each time is time-consuming.

Accordingly, Embodiment 2 is a specific example of Embodiment 1 described above, and will be described below to solve at least part of the above-described problems. FIG. 3 is a block diagram showing the overall configuration of the cloud system 1000 according to the second embodiment. A cloud system 1000 includes a computer environment 20 and a container placement device 30 .

A computer environment 20 includes a plurality of instances 21, 22, . . . 2n (n is a natural number). Here, each of the instances 21 to 2n is, for example, a physical or virtual server device. Also, each instance is partitioned into one or more containers, and instance resources are allocated to each container so that an application that performs processing corresponding to a service can be independently executed on a container-by-container basis. For example, the instance 21 is partitioned into containers 211, 212, . Similarly, the instance 22 is partitioned into containers 221, 222, . Also, the instance 2n is partitioned into containers 2n1, 2n2, .

The computer environment 20 is a server-side environment of an IoT system that provides multiple services. A plurality of services are provided to different types of business such as farmland, stores such as convenience stores, and air conditioners such as buildings. For example, measured values measured by sensors (IoT devices) installed in farmlands, stores, and air conditioning equipment are collected in the computer environment 20 via the network N. Since each service is in a different industry, it is assumed that the time zones and seasons in which loads are concentrated are different, and the timing and amount of concentration of required resources are also different. In other words, it can be said that each service has different characteristics in the time-series variation of the data processing load.

Each of the instances 21 to 2n and the container placement device 30 are connected via a network N. The network N is a communication network such as the Internet or a LAN (Local Area Network).

The container placement device 30 is an example of the information processing device 10 described above, and can be said to be a container placement system realized by one or more information processing devices. FIG. 4 is a block diagram showing the configuration of the container placement device 30 according to the second embodiment. The container placement device 30 includes a storage device 31 , a control section 32 , a memory 33 and a communication section 34 .

The storage device 31 is a storage device such as a hard disk or flash memory. The storage device 31 stores setting information 311 , placement information 312 , measured values 313 , predicted values 314 , residuals 315 and a container placement program 316 .

The setting information 311 is various setting values used for performing container placement processing according to this embodiment. The setting information 311 includes, for example, the upper limit of each resource usage, the collection interval for collecting measured values, the container placement interval, the unit and cycle used for time series analysis, the threshold for determining the presence or absence of regularity, and the Contains the interval for reviewing the forecast model. Here, the resource usage types (measurement items) include, but are not limited to, CPU, memory, disk I/O, and network bandwidth (communication speed, etc.). The threshold for determining the presence or absence of regularity can also be said to be the threshold for determining that the predictive model is no longer applicable. Also, the operator of the computer environment 20 preliminarily sets (stores) the setting information 311 in the storage device 31 .

The arrangement information 312 is information indicating containers and services arranged in each instance in the computer environment 20, and is, for example, a combination of an instance ID, a container ID, and a service (tenant) ID. The placement information 312 is the regularly collected container placement during actual operation. It should be noted that the format of the layout candidate, which will be described later, may correspond to the layout information 312 .

The measured value 313 is a set of pairs of measured values and measurement times for a plurality of measurement items for each container. Forecasts 314 are a set of forecasts at future times for resource usage corresponding to the metrics in measurements 313 for each container. The predicted value 314 also includes a predicted value of load for each service provided in the entire computer environment 20 . The residual 315 is a set of difference values between the corresponding container, measurement item, and (measured/predicted) time between the measured value 313 and the predicted value 314 .

The container placement program 316 is a computer program in which processing of the container placement method according to the second embodiment is implemented.

The memory 33 is, for example, a volatile storage device such as a RAM (Random Access Memory). The communication unit 34 is an interface for communicating with the outside of the container placement device 30 .

The control unit 32 is a processor, such as a CPU, that controls each component of the container placement device 30 . The control unit 32 loads the container placement program 316 from the storage device 31 into the memory 33 and executes the container placement program 316 . Thereby, the control unit 32 implements the functions of the collection unit 321 , the time series analysis unit 322 , the prediction unit 323 , the placement candidate calculation unit 324 and the rearrangement unit 325 .

The collection unit 321 is an example of the collection unit 11 described above, collects the measured values 313 from each of the services 2111 and the like in operation at each collection interval, and stores them in the storage device 31 .

The time-series analysis unit 322 is an example of the time-series analysis unit 12 described above, and performs time-series analysis processing on the measured value 313 to analyze trend data, seasonal data, and irregular data. and Trend fluctuation data is a set of measured values that indicate the trend of change along the time series, such as monotonous increase, monotonous decrease, periodic fluctuation, and constant value (no fluctuation) in a predetermined period of time series data in a specific period. Seasonal variation data is a collection of measured values that show a trend of change as a certain degree of long-term periodicity (regularity), such as an increase or decrease in data that occurs seasonally. Note that trend fluctuation data and seasonal fluctuation data are examples of regular fluctuation data. Irregular data is a set of measurements taken during periods in which no regularity of trend or seasonal variation was found. In other words, the irregular fluctuation data is a set of measured values other than the trend fluctuation data and the seasonal fluctuation data among the time-series data of the measured values 313 . Here, as a method of decomposing time series data into trend, seasonal and irregular fluctuation data, for example, the stl function of R language (https://www.r-project.org/) for statistical development can be cited. but not limited to.

Also, the time-series analysis unit 322 executes the time-series analysis process for the measured value 313 when the time-series fluctuation tendency of the measured value 313 and the predicted value 314 satisfies a predetermined condition. As a result, time-series analysis processing is not executed each time measurement values are collected, and the cost of calculation processing in the container placement device 30 can be reduced. At the same time, calculation processing of arrangement candidates after analysis processing and rearrangement processing are also avoided.

In particular, the time series analysis unit 322 calculates the difference between the measured value 313 and the predicted value 314 at corresponding times, and if there is regularity in the fluctuation trend in the time series of the difference, the time series analysis process is performed. . As a result, unknown fluctuation components that were initially regarded as irregular fluctuations and were not taken into account at the time of calculation of the predicted values are detected as new regularities, and the new regularities are further taken into account to make predictions. Can improve accuracy. In particular, even if the number of measurement data is small, the possibility of detecting regularity in irregular fluctuation data more quickly increases.

The prediction unit 323 is an example of the prediction unit 13 described above, and uses the trend change data and seasonal change data decomposed by the time series analysis unit 322 to predict each service for a predetermined period from the current time to the next collection time. Anticipate load. The predetermined period may be, for example, a period from the current collection time to the time when the collection interval included in the setting information 311 has passed (the time of the next collection) or longer. Here, in particular, the prediction unit 323 calculates a predicted load value for each service provided in the entire computer environment 20 . Also, the prediction unit 323 may learn, for example, a prediction model that calculates a predicted value of resource usage in a combination of instances, containers, and services from trend variation data and seasonal variation data. Then, the prediction unit 323 uses the learned prediction model to calculate the predicted value of the resource usage of each container corresponding to the current container arrangement, adds up these predicted values for each service, You can ask for the load.

The placement candidate calculation unit 324 is an example of the above-described placement candidate calculation unit 14, and calculates allocation of instances, containers, and services, that is, container placement candidates, based on the load prediction values calculated by the prediction unit 323. . For example, the placement candidate calculation unit 324 calculates a plurality of placement candidates that satisfy the predicted value of the load, and selects a placement candidate to be rearranged from among the calculated plurality of placement candidates based on a predetermined priority condition. select. Here, the priority condition is a condition for assigning a priority to be selected to each placement candidate, and is stored in the storage device 31 in advance. The priority condition is, for example, to give a higher priority to a container having a smaller estimated value of resource usage according to the arrangement state of containers. Thus, placement candidates optimized to meet the predicted future load can be obtained.

More specifically, the placement candidate calculation unit 324 calculates an individual predicted value of resource usage for each container for each of the plurality of placement candidates. Then, the placement candidate calculation unit 324 determines that the total value of the individual prediction values does not exceed the upper limit of the resource usage in the computer environment 20 as higher priority. Then, the placement candidate calculation unit 324 selects the one with the highest priority as a placement candidate to be rearranged. At this time, the placement candidate calculation unit 324 may use the above-described learned prediction model to calculate the predicted value of the resource usage amount for the set of instance, container, and service in the placement candidate as an individual predicted value. This increases the probability of finding placement candidates that are optimized to meet the predicted future load.

Further, when there is no placement candidate whose total value does not exceed the upper limit value among the plurality of placement candidates, the placement candidate calculation unit 324 determines that the product of the number of instances added within a predetermined period and the addition time is the minimum. It is preferable to select a placement candidate that is the target of rearrangement.

Further, when there are a plurality of placement candidates with the highest priority based on the predetermined priority condition, the placement candidate calculation unit 324 selects an instance among the plurality of placement candidates with the highest priority. It is desirable to select a placement candidate to be relocated based on the total value of predicted values of resource usage for each.

Further, the arrangement candidate calculation unit 324 calculates arrangement candidates by allocating each of the plurality of services to one of the plurality of containers.

The rearrangement unit 325 is an example of the rearrangement unit 15 described above, and rearranges the containers in the computer environment 20 based on the placement candidates selected as the rearrangement targets by the placement candidate calculation unit 324 .

FIG. 5 is a flow chart showing the flow of the container arrangement method according to the second embodiment. First, the collection unit 321 refers to the storage device 31 and determines whether or not the collection interval included in the setting information 311 has arrived (S201). If not, it waits for a certain period of time (S202), and executes step S201 again.

When it is determined in step S201 that the collection interval has arrived, the collection unit 321 collects the allocation information 312 and the resource usage measurement values 313 of all containers in the computer environment 20, and stores the allocation information 312 and the measurement values 313 in the storage device. 31 (S203).

Next, the time-series analysis unit 322 calculates a residual 315 between the current measured value 313 and the last calculated predicted value 314, and stores it in the storage device 31 (S204). That is, the time-series analysis unit 322 identifies, from the storage device 31, the predicted value 314 associated with the same predicted time as the measurement time of the (current) measured value 313 collected in step S203. Then, the time-series analysis unit 322 calculates, as a residual 315, the difference between the current measured value 313 and the specified predicted value 314 corresponding to the type of container and resource.

Then, the time-series analysis unit 322 analyzes the time-series fluctuation trend of the residual 315 (S205). For example, the time series analysis unit 322 analyzes the variation pattern of the residual 315 by arbitrary statistical analysis processing.

Subsequently, the time-series analysis unit 322 determines whether or not there is regularity in the time-series fluctuation trend of the residual 315 (S206). For example, if the residual 315 continues to be a constant value, or if the fluctuation pattern of the residual 315 is monotonically increasing, monotonically decreasing, or periodic, it is determined that there is regularity. Further, the above-described “threshold for determining the presence or absence of regularity” included in the setting information 311 may be used for determination of regularity.

If it is determined in step S206 that there is regularity, the time-series analysis unit 322 reads the measured value 313 in the period corresponding to the “unit and cycle used for time-series analysis” included in the setting information 311 from the storage device 31. , time-series analysis is performed (S207). The time-series analysis unit 322 decomposes the measured values 313 of the read period into trend fluctuation data, seasonal fluctuation data, and irregular fluctuation data by time-series analysis based on the "unit and period used for time-series analysis". . Then, the time-series analysis unit 322 stores the decomposed trend fluctuation data, seasonal fluctuation data, and irregular fluctuation data in the storage device 31 . It should be noted that if the "forced prediction model review interval" included in the setting information 311 has arrived, it is determined as YES in step S206. Also, in the "first time" of the loop, the container arrangement method is determined to be YES in step S206.

In the time-series analysis process in step S207, it is determined in step S206 immediately before that there is regularity in the trend of variation in the time-series of residuals. Therefore, in the previous time-series analysis process, there is a high possibility that the components that were regarded as irregular fluctuation data are regarded as trend fluctuation or seasonal fluctuation data. In other words, there is a case where the regularity that could not be detected until the previous time is newly detected. Then, the prediction accuracy can be improved by learning the prediction model in a state where the data in which the newly detected regularity is included in the trend fluctuation or seasonal fluctuation data. And since step S207 is performed only when it is YES in step S206, calculation cost can be reduced.

After that, the prediction unit 323 uses the trend change data and the seasonal change data decomposed in step S207 to calculate the predicted value of resource usage (predicted value of load) for each service from the present to the next predetermined period ( S208).

Then, the placement candidate calculation unit 324 calculates placement candidates from the predicted values calculated in step S208 (S209). After that, the placement candidate calculation unit 324 selects a placement candidate from among the placement candidates based on the priority condition (S210).

A specific processing example of steps S209 and S210 will be described below. First, the placement candidate calculation unit 324 derives a set (placement candidate) of pairs of instances, containers, and services from the number of instances in the computer environment 20, the number of containers that can be assigned to each instance, and the type of service to be provided. Then, the placement candidate calculation unit 324 calculates an individual predicted value of resource usage for each pair in each placement candidate using the prediction model described above. For example, the placement candidate calculation unit 324 obtains an individual prediction value by simulating each placement candidate using a prediction model. Then, the placement candidate calculation unit 324 calculates the total value of the individual prediction values for each placement candidate. The placement candidate calculation unit 324 adds a placement candidate whose total value satisfies the load prediction value to the placement candidate group. Therefore, the placement candidate group may include a plurality of placement candidates.

In this case, the placement candidate calculation unit 324 assigns priority to each placement candidate in the placement candidate group based on the priority condition. First, the placement candidate calculation unit 324 determines whether or not the individual prediction value in the placement candidate exceeds the upper limit value of each resource usage included in the setting information 311 . If none of the individual prediction values in the placement candidate exceeds the upper limit, the placement candidate calculation unit 324 gives the placement candidate the highest priority. On the other hand, if some or all of the individual prediction values in the placement candidate exceed the upper limit, the placement candidate calculation unit 324 adds given priority. For example, the placement candidate calculation unit 324 acquires the number of instances to be added and the time to add (additional time) according to the simulation, and calculates the product of the number of instances to be added and the additional time for each placement candidate. do. Then, the placement candidate calculation unit 324 assigns a higher priority to the larger product.

After assigning priority to all of the plurality of placement candidates, if the highest priority is one, the placement candidate calculation unit 324 selects the placement candidate to which the highest priority is assigned as a rearrangement target. , the process proceeds to step S211.

Further, when there are multiple highest priorities, the placement candidate calculation unit 324 calculates the total value of the individual prediction values of each instance in each placement candidate, and calculates the difference between the resource upper limit value of the instance and the total value. calculate. Then, the placement candidate calculation unit 324 calculates the product of differences for each placement candidate, selects the placement candidate with the largest product of differences as a rearrangement target, and proceeds to step S211. In other words, among the placement candidates within the resource upper limit, the one with the least variation in resource usage rate between instances is selected.

Also, if there is no placement candidate given the highest priority, the placement candidate calculation unit 324 selects the placement candidate given the highest priority among the given priorities as a rearrangement target. and proceed to step S211. That is, from among the placement candidates exceeding the resource upper limit, the one with the largest product of the number of added instances and the added time is selected.

After step S210, the rearrangement unit 325 rearranges the containers and services to each instance in the computer environment 20 based on the selected placement candidates (S211). After step S210, the time-series analysis unit 322 calculates the resource usage forecast value for the set of the instance, container, and service in the selected placement candidate using the forecast model, and associates the forecast time with the forecast value. may be stored in the storage device 31. Then, these may be used as the previous predicted values in the next step S204.

After NO in step S206 or step S211, the process advances to step S202 to repeat the container arrangement method.

As described above, according to the present embodiment, for example, it is possible to optimize the arrangement of containers in a service using container-type virtualization targeting IoT. That is, in this embodiment, the time-series resource usage status of each container is monitored, and the optimum combination of container placement is obtained based on the trend and seasonal fluctuations in the resource usage status. As a result, container placement that optimizes resource usage such as CPU usage and memory usage can be realized. In addition, by performing seasonal variation analysis only at the timing when the trend changes and at user-designated intervals, the rearrangement can be performed quickly.

<Other embodiments>
In the above-described embodiment, the hardware configuration is described, but the configuration is not limited to this. The present disclosure can also implement arbitrary processing by causing a CPU to execute a computer program.

In the above examples, the programs can be stored and delivered to computers using various types of non-transitory computer readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (eg, flexible discs, magnetic tapes, hard disk drives), magneto-optical recording media (eg, magneto-optical discs), CD-ROMs (Read Only Memory), CD-Rs, CD-R/W, DVD (Digital Versatile Disc), semiconductor memory (eg, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). The program may also be delivered to the computer on various types of transitory computer readable medium. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. Transitory computer-readable media can deliver the program to the computer via wired channels, such as wires and optical fibers, or wireless channels.

It should be noted that the present disclosure is not limited to the above embodiments, and can be modified as appropriate without departing from the spirit of the present disclosure. In addition, the present disclosure may be implemented by appropriately combining each embodiment.

Some or all of the above embodiments may also be described in the following additional remarks, but are not limited to the following.
(Appendix A1)
a collection unit that periodically collects measured values of resource usage from a computer environment in which a plurality of containers are arranged;
a time-series analysis unit that decomposes the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing;
a prediction unit that calculates a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a placement candidate calculation unit that calculates a placement candidate for containers in the computer environment based on the predicted value;
a rearrangement unit that rearranges containers in the computer environment based on the placement candidates;
Information processing device.
(Appendix A2)
The time series analysis unit
The information processing apparatus according to appendix A1, wherein the time-series analysis process is executed when the fluctuation tendency of the measured value and the predicted value in the time series satisfies a predetermined condition.
(Appendix A3)
The time series analysis unit
calculating a difference between the measured value and the predicted value at corresponding times;
The information processing apparatus according to appendix A2, wherein the time-series analysis process is performed when there is regularity in the fluctuation trend in the time-series of the differences.
(Appendix A4)
The placement candidate calculation unit
calculating a plurality of placement candidates;
selecting a placement candidate to be rearranged from among the plurality of calculated placement candidates based on a predetermined priority condition;
The information processing apparatus according to any one of Appendixes A1 to A3, wherein the rearrangement unit performs the rearrangement based on the selected arrangement candidate.
(Appendix A5)
The placement candidate calculation unit
calculating an individual predicted value of resource usage for each container for each of the plurality of placement candidates;
determining that the total value of the individual prediction values does not exceed the upper limit of resource usage in the computer environment as a higher priority;
The information processing apparatus according to appendix A4, wherein the one with the highest priority is selected as a placement candidate for the rearrangement target.
(Appendix A6)
The placement candidate calculation unit
If none of the plurality of placement candidates has a total value that does not exceed the upper limit, the placement candidate that minimizes the product of the number of instances added within the predetermined period and the additional time is rearranged. The information processing apparatus according to appendix A5, which is selected as a target.
(Appendix A7)
The placement candidate calculation unit
when there are a plurality of placement candidates with the highest priority based on the predetermined priority condition, a predicted resource usage amount for each instance among the plurality of placement candidates with the highest priority; The information processing apparatus according to any one of appendices A4 to A6, wherein the arrangement candidate for the rearrangement is selected based on the total value of .
(Appendix A8)
The placement candidate calculation unit
The information processing apparatus according to any one of Appendixes A1 to A7, wherein each of a plurality of services is assigned to one of the plurality of containers to calculate the placement candidate.
(Appendix B1)
the computer
Periodically collect resource usage measurements from a multi-container computing environment,
Decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing,
calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
calculating placement candidates for the container in the computer environment based on the predicted value;
A container placement method for rearranging containers in the computer environment based on the placement candidates.
(Appendix C1)
a process of periodically collecting resource usage measurements from a computer environment in which a plurality of containers are arranged;
a process of decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis;
a process of calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a process of calculating placement candidates for containers in the computer environment based on the predicted values;
a process of rearranging containers in the computer environment based on the placement candidates;
A container placement program that causes a computer to run a .

Although the present invention has been described with reference to the embodiments (and examples), the present invention is not limited to the above-described embodiments (and examples). Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

10 information processing device 11 collection unit 12 time series analysis unit 13 prediction unit 14 placement candidate calculation unit 15 rearrangement unit 1000 cloud system 20 computer environment 21 instance 211 container 2111 service 212 container 2121 service 22 instance 221 container 2211 service 222 container 2221 service 2n instance 2n1 container 2n11 service 2n2 container 2n21 service N network 30 container placement device 31 storage device 311 setting information 312 placement information 313 measured value 314 prediction value 315 residual 316 container placement program 32 control unit 321 collection unit 322 time series analysis unit 323 Prediction unit 324 Placement candidate calculation unit 325 Rearrangement unit 33 Memory 34 Communication unit

Claims (9)

a collection unit that periodically collects measured values of resource usage from a computer environment in which a plurality of containers are arranged;
a time-series analysis unit that decomposes the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing;
a prediction unit that calculates a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a placement candidate calculation unit that calculates a placement candidate for containers in the computer environment based on the predicted value;
a rearrangement unit that rearranges containers in the computer environment based on the placement candidates;
with
The time series analysis unit
When the trend of variation in the time series of the measured values and the predicted values satisfies a predetermined condition, the time series analysis process is further executed.
Information processing equipment. The time series analysis unit
calculating a difference between the measured value and the predicted value at corresponding times;
The information processing apparatus according to claim 1 , wherein the time-series analysis process is performed when there is regularity in the fluctuation tendency in the time-series of the differences. The placement candidate calculation unit
calculating a plurality of placement candidates;
selecting a placement candidate to be rearranged from among the plurality of calculated placement candidates based on a predetermined priority condition;
The information processing apparatus according to Claim 1 or 2 , wherein the rearrangement unit performs the rearrangement based on the selected arrangement candidate. The placement candidate calculation unit
calculating an individual predicted value of resource usage for each container for each of the plurality of placement candidates;
determining that the total value of the individual prediction values does not exceed the upper limit of resource usage in the computer environment as a higher priority;
The information processing apparatus according to claim 3 , wherein the one with the highest priority is selected as a placement candidate for the rearrangement. The placement candidate calculation unit
If none of the plurality of placement candidates has a total value that does not exceed the upper limit, the placement candidate that minimizes the product of the number of instances added within the predetermined period and the additional time is rearranged. The information processing apparatus according to claim 4 , which is selected as an object. The placement candidate calculation unit
when there are a plurality of placement candidates with the highest priority based on the predetermined priority condition, a predicted resource usage amount for each instance among the plurality of placement candidates with the highest priority; 6. The information processing apparatus according to any one of claims 3 to 5 , wherein the arrangement candidate for the rearrangement is selected based on the total value of . The placement candidate calculation unit
The information processing apparatus according to any one of claims 1 to 6 , wherein each of a plurality of services is assigned to one of the plurality of containers to calculate the placement candidate. the computer
Periodically collect resource usage measurements from a multi-container computing environment,
Decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis processing,
calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
calculating placement candidates for the container in the computer environment based on the predicted value;
rearranging the containers in the computer environment based on the placement candidates ;
When the trend of variation in the time series of the measured values and the predicted values satisfies a predetermined condition, the time series analysis process is further executed.
Container placement method. a process of periodically collecting resource usage measurements from a computer environment in which a plurality of containers are arranged;
a process of decomposing the measured values into regular fluctuation data and irregular fluctuation data by time-series analysis;
a process of calculating a predicted value of resource usage in the computer environment from now until a predetermined period from now based on the rule fluctuation data;
a process of calculating placement candidates for containers in the computer environment based on the predicted values;
a process of rearranging containers in the computer environment based on the placement candidates;
a process of further executing a process of analyzing the time series when the trend of variation in the time series of the measured values and the predicted values satisfies a predetermined condition;
A container placement program that causes a computer to run a .

Images