KR102561320B1 - A container replica recommendation system through resource trend prediction and a recommendation method - Google Patents

Abstract

A container replica recommendation system according to an embodiment of the present invention is a container replica recommendation system for operating and managing a cloud server that provides container-based cloud services, comprising: a collection module that collects past usage information, which is information about resources of a physical server used when a container is used; an analysis module processing the past usage information by a predetermined processing method to calculate trend information, which is resource tendency; and a recommendation module that calculates an initial range, that is, a range of the number of pods that can currently be operated on any worker node, based on the trend information.

Description

Container replica recommendation system and recommendation method through resource trend prediction {A CONTAINER REPLICA RECOMMENDATION SYSTEM THROUGH RESOURCE TREND PREDICTION AND A RECOMMENDATION METHOD}

The present invention relates to a replica recommendation system and method, and more particularly, to a recommendation system and method for operating a cloud by predicting a resource trend and recommending a container replica.

As the technological power of Internet networks increases and the spatial limitations of computer work such as working from home disappear, the cloud environment is rapidly growing. Cloud services are largely divided into a method of providing a virtualized virtual machine and a method of providing through a container. Container-based cloud services are gaining popularity due to user convenience and simplified maintenance.

Here, it is very important how to manage the worker node on the physical machine in order to provide the cloud service without interruption from the point of view of the operator providing the Wood service. In the past, the operation of the worker node is set in advance by specifying the pod running on the worker node in advance. This method does not adapt to the changing resources in real time, and causes problems in that available resources of the physical server are wasted and inability to respond to rapidly increasing usage.

SUMMARY OF THE INVENTION The present invention is intended to solve the above problems, and is intended to provide a container replica recommendation system and container replica recommendation method capable of efficiently managing resources by predicting resource trends and effectively responding to situations.

A container replica recommendation system according to an embodiment of the present invention is a container replica recommendation system for operating and managing a cloud server that provides container-based cloud services, comprising: a collection module that collects past usage information, which is information about resources of a physical server used when a container is used; an analysis module processing the past usage information by a predetermined processing method to calculate trend information, which is resource tendency; and a recommendation module that calculates an initial range, that is, a range of the number of pods that can currently be operated on any worker node, based on the trend information.

In addition, the predetermined processing method may be a method of separating the past usage information into trend information, seasonality information obtained by analyzing a certain pattern according to a date, and irregular information not having a certain pattern.

Further, the present invention further includes a prediction module that calculates predicted usage information, which is information on resources of the physical server predicted to be used during a predetermined future period, based on the trend information and current usage information, which is information on resources of the currently operating physical server, wherein the recommendation module may recommend a range of replicas based on the predicted usage information.

In addition, the recommendation module recommends a first initial range when the resources of the currently operating physical server are less than or equal to a predetermined ratio of the available resources of the physical server, and recommends a second initial range when the resources of the currently operating physical server exceed a predetermined ratio of the available resources of the physical server.

In addition, a control module for controlling worker nodes including containers so that the number of pods is maintained within an initial range, wherein the prediction module calculates predicted usage information, which is information about resources of the physical server predicted to be used in a period after the predetermined future period, and the control module can increase or decrease the number of pods that can be operated on any worker node according to the increase or decrease of the predicted usage information.

The preference evaluation module may further include a preference evaluation module that calculates a preference evaluation value related to management preferences of the manager of the cloud server by a predetermined preference calculation method, wherein the collection module collects operation information, which is information generated while the physical server is operated by a manager, and the predetermined preference calculation method is a method of calculating the preference evaluation value based on the operation information, and the recommendation module may calculate a correction range by modifying the initial range in consideration of the preference evaluation value.

In addition, the preference evaluation value may be calculated in consideration of safety and efficiency aspects.

In addition, the collection module collects personal information of a manager when the operation information is not collected by a predetermined collection amount, and the predetermined preference calculation method considers the personal information and the operation information together. It may be a method of calculating the preference evaluation value.

In addition, it may further include a proficiency determination module for determining a manager's operational proficiency of the cloud server based on the operation information, and the predetermined preference calculation method may be a method of calculating the preference evaluation value by changing a weight to which the personal information and the operation information are applied based on the operation proficiency.

A container replica recommendation method according to an embodiment of the present invention is implemented by a container replica recommendation system and is implemented by a container replica recommendation system for operating and managing a cloud server providing a container-based cloud service. A container replica recommendation method comprising the steps of: collecting, by a collection module, past usage information, which is information about resources of a physical server used when a container is used; Calculating, by an analysis module, trend information that is a resource tendency from the past usage information using a predetermined processing method; and calculating, by a recommendation module, an initial range, which is a range of the number of pods that can currently be operated on a certain worker node, based on the trend information.

The container-based container replica recommendation system and container replica recommendation method using resource trend prediction according to the present invention can maximize the stability of cloud services.

In addition, manpower for server management can be minimized.

In addition, time required for server management can be minimized.

In addition, it is possible to maximize the efficiency of available resources of the cloud server.

However, the effects of the present invention are not limited to the above-mentioned effects, and effects not mentioned will be clearly understood by those skilled in the art from this specification and the accompanying drawings.

1 is a relationship diagram of a container replica recommendation system according to an embodiment of the present invention.
2 is a block diagram of a container replica recommendation system according to an embodiment of the present invention.
3 is a structural relationship diagram of a container replica recommendation system according to an embodiment of the present invention.
4 is a flowchart of a container replica recommendation method according to an embodiment of the present invention.
5 is a diagram for explaining a process of calculating trend information in a container replica recommendation method according to an embodiment of the present invention.
6 is a diagram for explaining a process of calculating an initial range and a modified range in a container replica recommendation method according to an embodiment of the present invention.
7 is a diagram for explaining a process of calculating a preference evaluation value by utilizing personal information in a container replica recommendation method according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings. However, the spirit of the present invention is not limited to the presented embodiments, and those skilled in the art who understand the spirit of the present invention may easily suggest other degenerative inventions or other embodiments included within the scope of the present invention through the addition, change, deletion, etc. of other components within the scope of the same idea, but this will also be said to be included within the scope of the present invention.

In addition, components having the same function within the scope of the same idea appearing in the drawings of each embodiment are described using the same reference numerals.

1 is a relationship diagram of a container replica recommendation system according to an embodiment of the present invention.

1 is a server management system according to an embodiment of the present invention may be a system that operates, maintains, and manages a physical server that provides a container-based cloud service, and operates, manages, creates, and coordinates container operations.

As a specific example, the container replica recommendation system 100 can monitor whether a problem occurs with a container, and if a problem occurs with a worker node that includes a container or a master node that can control the worker node, the problem can be resolved.

In addition, the container replica recommendation system 100 can effectively set and maintain the number of pods in a worker node in consideration of the resources of the physical server.

The container replica recommendation system system 100 may be connected to the physical server 200 and/or the external server 300 through a wired/wireless network. The server operating system may collect and receive all information generated while the container operates on the physical server, all information generated while the physical server operates, and/or necessary information from external servers.

The network in the present invention may be a core network integrated with a wired public network, a wireless mobile communication network, or a mobile Internet, and may refer to a worldwide open computer network structure that provides TCP/IP protocol and various services existing in the upper layer, that is, HTTP (Hyper Text Transfer Protocol), HTTPS (Hyper Text Transfer Protocol Secure), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), etc. It means a comprehensive data communication network that can do.

The physical server in the present invention may include other configurations for performing the server environment of the server. The server may include any type of device.

As an example, the server may be a digital device, such as a laptop computer, a notebook computer, a desktop computer, a web pad, or a mobile phone, which includes a processor and a memory and arithmetic capability.

As an example, the server may be a web server. However, it is not limited to this, and the type of server can be variously changed at a level obvious to those skilled in the art.

2 is a configuration diagram of a container replica recommendation system according to an embodiment of the present invention, and FIG. 3 is a configuration diagram of a container replica recommendation system according to an embodiment of the present invention.

Referring to FIGS. 2 and 3 , in the container replica recommendation system 100 according to an embodiment of the present invention, in the container replica recommendation system 100 for operating and managing a cloud server providing container-based cloud services, a collection module 110 that collects past use information, which is information about resources of physical servers used by containers, an analysis module 120 that processes the past use information by a predetermined processing method to calculate trend information that is a resource tendency, and the trend information It may include a recommendation module 130 that calculates an initial range, which is a range of the number of pods that can currently be operated on an arbitrary worker node, based on .

In addition, the container replica recommendation system 100 according to an embodiment of the present invention may further include a prediction module 140 that calculates predicted usage information, which is information about resources of the physical server predicted to be used during a predetermined future period, based on the trend information and current usage information, which is information about resources of the physical server currently in operation, and a preference evaluation module 150 that calculates a preference evaluation value related to management preferences of the manager of the cloud server by a predetermined preference calculation method.

In addition, the container replica recommendation system 100 according to an embodiment of the present invention may further include a proficiency determination module 160 for determining a manager's proficiency in operating the cloud server based on the operation information, and a storage module 170 for storing all information required to implement the container replica recommendation method.

In addition, the container replica recommendation system 100 according to an embodiment of the present invention may further include an interface module 190 that calculates an interface for displaying and receiving information necessary for system administrators and cloud service users to operate the container replica recommendation system 100, and a control module 180 that controls and controls worker nodes in the physical server to effectively provide container-based cloud services.

The collection module 110 may collect operation information, which is information generated when a manager manages a physical server and operates and controls a container.

The operation information may include both log information and/or metric data of the physical server/worker node. In addition, the operation information may include control data by which the manager controls the physical server and all information about the current operation status of the physical server and the container when the control data is generated.

For example, the operation information may include methods by which the manager controls and controls the physical server and/or the master node (container) in order to solve an abnormal situation in the current state of the physical server and container.

The collection module 110 may collect manager's personal information when the amount of operation information is not collected in advance.

As an example, the predetermined collection amount may be 1 TB. However, it is not limited to this, and the predetermined collection amount can be variously modified at a level obvious to those skilled in the art.

Personal information may include personal information of the manager and information generated while the manager controls the electronic device from the outside, regardless of the server operating system 100 .

To this end, the collection module 110 may collect necessary information by being network-connected to an external server, a public institution server, an open server, and the like.

For this purpose, prior approval may be obtained from the administrator.

For example, personal information may include navigation route selection information of a manager.

When searching for a navigation route, various routes may be recommended, and the navigation user may select one route among the multiple routes according to his or her preference.

For this reason, using the information about the manager's use of the navigation application can help determine the manager's preference.

For example, personal information may include vehicle control information of a manager.

Recently, various control information of automobiles is being collected based on the expansion of IOT technology and the trend of electronic devices of automobiles. The collection modules 110 and 111 may collect the information that the manager controls and controls the car from the car manufacturing server and/or the communication company server.

How a car is controlled can be an important piece of information in determining whether a manager is a person who values safety or efficiency.

The vehicle control information is fused together with the navigation information in the vehicle, and may be important information in determining how the manager controls the vehicle in a traffic situation and location.

For example, personal information may include information about the manager's age, gender, school of origin, major information, and region of origin.

As people get older, they may tend to value stability more than efficiency.

Women more than men may tend to value stability more than efficiency.

Because the academic traditions of universities affect students, the importance of efficiency and stability may differ depending on the school you are from.

Since the major you learn affects the learner, the importance of efficiency and stability may differ depending on the major.

Depending on whether the area of origin is an urban, rural, or fishing village affects residents, the importance of efficiency and stability may differ.

For example, if you live in an urban area, you may value efficiency more than stability. In addition, residents of rural/fishing villages may value stability more than efficiency.

The collection module 110 may collect past usage information, which is information about the resources of the physical server in which the container is used.

The resources of the physical server may mean physical server resources such as CPU, RAM, and storage.

Past use information collected by the collection module 110 may be stored in the storage module 170 .

The collection module 110 may collect current usage information, which is information about resources of a physical server in which containers are operated and used, in real time.

The analysis module 120 may process past use information by a predetermined processing method to calculate trend information, which is a tendency of a resource.

The predetermined processing method may be a method of separating the past use information into trend information, seasonality information obtained by analyzing a certain pattern according to a date, and irregular information not having a certain pattern.

A detailed description of this will be described later.

Trend information may be data sorted based on date and time.

This means that periodicity and pattern may occur in the use of data according to date and time.

Through this, it is possible to grasp the trend in which the resources of the physical server currently in use will increase or decrease.

The prediction module 140 may calculate predicted usage information, which is information about resources of the physical server that are predicted to be used during a predetermined future period, based on the current usage information collected by the collection module 110 .

The prediction module 140 may calculate predicted usage information based on current usage information and trend information.

As a specific example, the prediction module 140 may calculate how resources used in the physical server will be in the future by merging trend information with resources currently used in the physical server.

That is, if it is assumed that the time point at which the predicted usage information is calculated is 15:00 on December 1, 2022, the trend information at 15:00 on December 1, 2022, among the past usage information for the used resource collected at 15:00 on December 1, 2022 Considering, the predicted usage information expected to occur after 15:00 on December 1, 2022 can be calculated.

Here, the prediction module 140 may newly calculate the predicted usage information in units of a predetermined future period.

Because of this, since future predicted use information is calculated by utilizing the resources of the physical server used in real time, the resources of the physical server can be efficiently managed.

As an example, the predetermined future period may be 'one day'.

However, the present invention is not limited thereto, and the specific period of the predetermined future period can be variously modified at a level obvious to those skilled in the art.

The preference evaluation module 150 may calculate a preference evaluation value related to the manager's management preference according to a predetermined preference calculation method.

The preference evaluation module 150 may store a pattern classification model for classifying motion information into predetermined operation patterns.

The preference evaluation module 150 may generate a pattern classification model by deep learning based on classification information obtained by classifying motion information corresponding to past motion information.

Deep learning may be performed using a back propagation algorithm, which is an algorithm for updating weights of a neural network using labeled data of an output layer, but is not limited thereto.

In addition, since the deep neural network and the back propagation algorithm are conventionally known, a detailed description thereof may be omitted.

For example, the preference evaluation value may be composed of values for efficiency and stability, and the unit may be %, and the sum of efficiency and stability may be 100.

However, it is not limited thereto, and the form of the preference evaluation value can be variously modified at a level obvious to those skilled in the art.

Operation patterns may be classified by pattern in terms of stability and efficiency.

For example, the first operating pattern may be a pattern with stability of 0% and efficiency of 100%, and the second operating pattern may be a pattern with stability of 2% and efficiency of 98%. In this way, operating patterns can be classified in terms of all stability and efficiency.

The predetermined preference calculation method may be a method of calculating the preference evaluation value by classifying an operating pattern of the manager's container replica recommendation system 100 based on the operation information.

The preference evaluation value can be calculated by considering the aspects of stability and efficiency.

The manager's management tendency can be effectively analyzed by the preference evaluation module 150 .

The predetermined preference calculation method may be a method of calculating the preference evaluation value by considering both the personal information and the operation information.

The preference evaluation module 150 may calculate a preference evaluation value by utilizing personal information together when motion information is not collected as much as a predetermined collection amount.

The preference evaluation module 150 may store a personal calculation model that calculates efficiency and stability through personal information.

The preference evaluation module 150 may calculate a personal calculation model by deep learning based on values classified into efficiency and stability for personal information corresponding to past personal information.

The deep learning algorithm utilizes a known technique, and a detailed description thereof may be omitted.

The preference evaluation module 150 may finally calculate a preference evaluation value by taking a weighted average of the first preference evaluation value calculated through the pattern classification model and the second preference evaluation value calculated by the individual calculation model.

The preference evaluation module 150 may calculate the preference evaluation value by changing the weights of the first preference evaluation value and the second preference evaluation value according to the amount of motion information collected until motion information is collected as much as the predetermined collection amount.

For example, as the amount of motion information collected increases, the first preference evaluation value may be more weighted than the second preference evaluation value, and an average may be calculated to calculate a final preference evaluation value.

If the information of the operation information is too small, the first preference evaluation value may not fully reflect the manager's propensity and preference.

In order to compensate for this, a final preference evaluation value may be calculated by further weighting the second preference evaluation value using personal information for which a considerable amount of data has already been accumulated.

As the motion information gradually increases, the weight of the first preference evaluation value may be increased according to the first ratio with respect to the increased amount, and the weight of the second preference evaluation value calculated using the personal information may be decreased.

The predetermined preference calculation method may be a method of calculating the preference evaluation value by changing a weight to which the personal information and the motion information are applied based on the manipulation proficiency.

In other words, in the predetermined preference calculation method, when the amount of motion information collected is equal to or greater than a predetermined reference value, the ratio of weight reduction of the second preference evaluation value according to the weight increase of the first preference evaluation value according to the manipulation proficiency may be a method of changing the weight at a ratio different from the first ratio.

As a specific example, when the amount of motion information collected is equal to or greater than a predetermined reference value (change reference value) and the manager's operational proficiency is greater than or equal to a predetermined proficiency, the preference evaluation module may calculate the final preference evaluation value by setting the ratio of weight reduction of the second preference evaluation value to a second ratio, which is greater than the first ratio, according to the increase in the weight of the first preference evaluation value.

In addition, when the amount of motion information collected is greater than or equal to a predetermined reference value and the manager's operational proficiency is less than the predetermined proficiency, the preference evaluation module may calculate a final preference evaluation value by setting a ratio in which the weight of the second preference evaluation value is reduced as the weight of the first preference evaluation value increases to a third ratio that is smaller than the first ratio.

This may be because if the manager's proficiency is low, the reliability of the information is somewhat low because the consistency of the operation information is somewhat low. Conversely, if the manager's proficiency is high, it may be because the information reliability is somewhat high because the consistency of the operation information is somewhat high.

If the manipulation proficiency is not calculated by the proficiency determination module 160 or 116, the preference evaluation module may calculate the final preference evaluation value by maintaining the first ratio.

The proficiency determination module 160 may calculate the manager's operational proficiency of the container replica recommendation system 100 based on the operation information.

To this end, a proficiency judgment model may be stored in the proficiency judgment module 160 .

The proficiency determination module 160 may generate a proficiency judgment model by performing deep learning on how the physical server and virtual machine were manipulated and evaluation scores thereof according to the past states of the physical server and virtual machine.

As a specific algorithm for deep learning utilizes a known technique, a detailed description thereof may be omitted.

The proficiency judgment model can calculate the operational proficiency of the manager.

However, the proficiency judgment model requires only the control information necessary to determine the manipulation proficiency among the motion information and the state information of the physical server and container (worker node) at the time of control.

The proficiency judgment model can calculate the manipulation proficiency only when the amount of data required to determine the manipulation proficiency is equal to or greater than a predetermined standard.

Here, the amount of data required to determine the manipulation proficiency may be smaller than the change reference value.

This may be to effectively operate the above-described algorithm.

The proficiency determination module 160 may transmit the calculated manipulation proficiency to the preference evaluation module 150 .

The final preference evaluation value calculated by the preference evaluation module 150 may be transmitted to the recommendation module 130 .

The recommendation module 130 may recommend an initial range, which is a range of the number of Pods that can be operated on an arbitrary worker node, to the manager.

A worker node includes at least one pod, and a pod includes at least one container, and may refer to a classification system that exists in terms of software.

Worker node is a well-known technology, and a detailed description thereof may be omitted.

The recommendation module 130 may recommend a range of replicas based on predicted usage information.

As a specific example, the recommendation module 130 adds or subtracts a predetermined value to the predicted usage information according to how much the resources of the currently operating physical server occupy in the available resources of the physical server, and thereby calculates allowable usage information, which is an allowable resource of the physical server generated by container operation.

The allowed usage information may include first allowed usage information calculated by adding or subtracting a first allowed value to predicted usage information and second allowed usage information calculated by adding or subtracting a second allowed value to predicted usage information.

The first allowable value may be smaller than the second allowable value.

For example, the first allowable value and the second allowable value may be set in advance.

When the resources of the currently operating physical server exceed a predetermined ratio of the available resources of the physical server, the recommendation module 130 may calculate the first allowed usage information by adding and subtracting the first allowed value from the predicted usage information.

Alternatively, when the resources of the currently operating physical server are less than or equal to a predetermined ratio of the available resources of the physical server, the recommendation module 130 may calculate the second allowed usage information by adding and subtracting the second allowed value from the predicted usage information.

The recommendation module 130 may calculate an initial range based on the allowed use information.

That is, the recommendation module 130 may calculate the second initial range by utilizing the first permitted use information, and may calculate the first initial range by utilizing the second permitted use information.

The recommendation module 130 may calculate an average of upper values of allowed use information and an average of lower values of allowed use information in units of a predetermined future period.

The recommendation module 130 calculates the range of the number of pods to be provided from the physical server in order for the worker node to operate stably based on the average of the upper values of the permitted usage information and the average of the lower values of the permitted usage information, and may calculate the initial range.

In other words, the recommendation module 130 may recommend the resource of the currently operating physical server as the first initial range when the resource of the currently operating physical server is less than or equal to a predetermined ratio of the available resources of the physical server, and may recommend the resource of the currently operating physical server as the second initial range when the resource of the currently operating physical server exceeds a predetermined ratio of the available resource of the physical server.

The recommendation module 130 may calculate a corrected range by correcting the initial range in consideration of the preference evaluation value.

The preference evaluation value can be calculated by considering the aspects of stability and efficiency.

The recommendation module 130 may calculate the corrected range by correcting the initial range according to the stability and efficiency of the preference evaluation value.

The recommendation module 130 may calculate the correction range according to the weight of stability and efficiency of the preference evaluation value within the initial range.

As a specific example, the recommendation module 130 may calculate the correction range biased toward the minimum value of the initial range when stability is higher than efficiency in the preference evaluation value.

Conversely, when efficiency is higher than stability in the preference evaluation value, the recommendation module 130 may calculate the correction range biased toward the maximum value of the initial range.

Through this, it is possible to recommend a range of replicas in consideration of the manager's management pattern and tendency.

The control module 180 may control and control worker nodes in the physical server in order to effectively provide container-based cloud services.

As an example, the control module 180 may be a master node.

However, it is not limited thereto, and the type of control module 180 can be variously modified at a level obvious to those skilled in the art.

The control module 180 may maintain pods in a worker node based on the number of replicas, that is, pods requested by the manager.

Replicas can mean the number of pods that a node needs to maintain.

The control module 180 may control worker nodes including containers so that the number of pods is maintained within the initial range.

The interface module 190 may calculate an interface for displaying and receiving information necessary for the system administrator and cloud service user to operate the container replica recommendation system 100 .

The interface module 190 may be connected to a manager's computing device and a user's computing device through a wired/wireless network to transmit the calculated interface and receive necessary information.

The interface module 190 may transmit an interface displaying an initial range and a modified range, which are replica ranges, to the manager computing device.

The interface module 190 may configure the interface to display that the correction range is within the range considering the manager's management propensity.

The storage module 170 may store all information required to implement the container replica recommendation method.

For example, the storage module may include internal memory and/or external memory.

For example, the built-in memory may include at least one of volatile memory (eg, DRAM, SRAM, or SDRAM), non-volatile memory (eg, one time programmable ROM (OTPROM), PROM, EPROM, EEPROM, mask ROM, flash ROM, flash memory, hard drive, or solid state drive (SSD)).

The external memory may include a flash drive, for example, a compact flash (CF), secure digital (SD), micro-SD, mini-SD, extreme digital (xD), multi-media card (MMC), or memory stick.

Hereinafter, a container replica recommendation method implemented by each module in the container replica recommendation system 100 will be described in detail.

4 is a flowchart of a container replica recommendation method according to an embodiment of the present invention.

Hereinafter, a detailed description may be omitted to the extent that it overlaps with the above description.

Referring to FIG. 4 , a container replica recommendation method according to an embodiment of the present invention is implemented by a container replica recommendation system and is implemented by a container replica recommendation system for operating and managing a cloud server providing a container-based cloud service. The step of collecting, by a collection module, past usage information, which is information about the resources of physical servers used by containers, the step of calculating, by an analysis module, trend information that is a resource tendency from the past usage information by a predetermined processing method, and the recommendation module, the trend information Based on this, it may include calculating an initial range, which is a range of the number of pods that can currently be operated on any worker node.

Operation information may be collected by the collection module.

If motion information is not collected as much as a predetermined collection amount, manager's personal information is collected, proficiency is determined, and a preference evaluation value is calculated, and the personal information and proficiency can be used.

If motion information is collected by a predetermined collection amount, a preference evaluation value may be calculated only with motion information.

In addition, the collection module may collect past usage information, and the analysis module may calculate trend information through a predetermined processing method.

The prediction module may calculate predicted usage information, which is information about physical server resources for a predetermined future period, based on the trend information and the current usage information.

The administrator may request to set a replica to the container replica recommendation system through his/her computing device.

When the interface module receives a replica request, the recommendation module may calculate an initial range and a modified range, calculate an interface in which the initial range and modified range are recommended, and transmit the calculated interface to the manager's computing device.

The manager can designate a replica within the scope of modification, or within the initial scope but not within the scope of modification.

The control module can control the worker node to maintain the number of pods according to the replica specified by the manager for a predetermined future period based on the replica specified by the manager.

The prediction module may calculate predicted use information, which is information about resources of the physical server predicted to be used in a period after the predetermined future period.

The control module can increase or decrease the number of pods that can be operated on any worker node according to the increase or decrease of predicted usage information.

After a predetermined future period (first predetermined future period, T10, see FIG. 6) has elapsed, the prediction module may further calculate second predicted usage information, which is information about resources of the physical server predicted to be used during the next predetermined future period (second predetermined future period, T20, see FIG. 6).

The control module may compare a first average, which is an average of first predicted usage information, which is predicted usage information for a first predetermined future period, and a second average, which is an average of second predicted usage information, which is predicted usage information, for a second predetermined future period, and if the first average is higher than the second average, set a replica lower than the replica specified by the manager to manage the worker node.

Alternatively, when the first average is lower than the second average, the control module may manage the worker node by setting the replica higher than the replica designated by the manager.

5 is a diagram for explaining a process of calculating trend information in a container replica recommendation method according to an embodiment of the present invention.

Referring to FIG. 5 (a) , past use information may be collected by the collection module.

For example, past usage information may be collected on a yearly basis as time-series information.

However, it is not limited to this, and the collection cycle of the past usage information can be variously modified at a level obvious to those skilled in the art.

Past use information is decomposed by a predetermined processing method, and trend information (see FIG. 5(b)) for confirming the trend of past use information, seasonality information (see FIG. 5(c)) for confirming the seasonality of past use information, and irregular information (see FIG. 5(d)) for confirming non-regular data among past use information can be calculated.

As an example, the predetermined processing method may refer to a multiple decomposition analysis method.

However, it is not limited thereto, and the predetermined processing method can be variously modified at a level obvious to those skilled in the art.

Trend information analyzed by the analysis module may be transmitted to the recommendation module.

6 is a diagram for explaining a process of calculating an initial range and a modified range in a container replica recommendation method according to an embodiment of the present invention.

Referring to FIG. 6 , the prediction module may calculate predicted use information F10, which is a resource of a physical server expected to be generated by a worker node during a predetermined future period T10.

Here, although FIG. 6 shows a graph in which the resources of the physical server are converted into numerical values from 0 to 100, the present invention is not limited thereto, and the unit representing the predicted usage information can be variously modified at a level obvious to those skilled in the art.

When the resources of the currently operating physical server exceed a predetermined ratio of the available resources of the physical server, the recommendation module adds and subtracts the first allowed value E10 to the predicted usage information F10 to calculate the first allowed usage information M11.

Here, the predetermined ratio may be 60%.

However, it is not limited thereto, and the predetermined ratio can be variously modified at a level obvious to those skilled in the art.

When the resources of the currently operating physical server are less than or equal to a predetermined ratio of the available resources of the physical server, the recommendation module adds and subtracts the second allowed value E20 to the predicted usage information F10 to calculate the second allowed usage information M12.

The recommendation module calculates the average of the upper limit value and the average of the lower limit value of the calculated first allowed use information M11 or the second allowed use information M12, and based on the calculated range, the initial range, which is a range of a certain number of pods in which the worker node can be stably operated, can be calculated.

The recommendation module may calculate the modified range by considering the initial range as the preference evaluation value.

As an example, in the method of calculating the correction range, the maximum value of the correction range can be calculated by adding a value obtained by multiplying the difference between the median and the maximum value by the ratio of the utility among the preferred evaluation values to the median value, which is the middle of the maximum and minimum values representing the initial range.

In addition, in the method of calculating the correction range, the value obtained by subtracting the value obtained by multiplying the difference between the median and the minimum value by the ratio of stability among the preferred evaluation values to the median value, which is the middle of the maximum and minimum values representing the initial range, can be calculated as the minimum value of the correction range.

For example, it may be assumed that the initial range (the first initial range or the second initial range) is 10 to 20, and the preferred evaluation values are 80 for efficiency and 20 for stability. In this case, the median value is '15', and '4', which is the value obtained by multiplying '5', which is the difference between the median value and the maximum value, by '0.8', which is the ratio of utility among the preferred evaluation values, is added to the median value '15'. '19', the maximum value of the correction range, can be calculated. In addition, '14', the maximum value of the correction range, can be calculated by subtracting '1', which is '0.2', which is the ratio of stability among the preference evaluation values, to '5', which is the difference between the median value and the minimum value, from the median value '15'. Through this process, the correction range can be calculated from '14' to '19'.

As described above, the initial range and the modified range reflecting the manager's management propensity are recommended to the manager together, so the manager can efficiently set the replica.

7 is a diagram for explaining a process of calculating a preference evaluation value by utilizing personal information in a container replica recommendation method according to an embodiment of the present invention.

Referring to FIG. 7 , for example, when motion information on the X axis is 100% utilized, it may mean when motion information is collected for a predetermined collection amount.

If no motion information is collected, only personal information is 100% utilized and a preference evaluation value can be calculated.

As the motion information gradually increases, the weight of the second preference evaluation value calculated by using the personal information according to the first ratio may decrease with respect to the increased amount.

However, the present invention is not limited thereto, and the first ratio can be variously modified at a level apparent to those skilled in the art.

7(a) is a diagram illustrating a case in which the manipulation proficiency is equal to or greater than a predetermined proficiency level. Referring to FIG. 7(a), a predetermined reference value X10 of the amount of operation information collected may be half of a predetermined collection amount.

However, it is not limited thereto, and the predetermined reference value can be variously modified at a level obvious to those skilled in the art.

When the amount of motion information collected is less than the predetermined reference value X10, the first preference evaluation value and the second preference evaluation value are adjusted according to the first ratio B11 to calculate the final preference evaluation value. However, when the amount of motion information collected is greater than or equal to the predetermined reference value, the first preference evaluation value and the second preference evaluation value may be adjusted according to the second ratio B12 to calculate the final preference evaluation value.

For this reason, for example, even if only 75% of the predetermined collection amount of motion information is collected, 100% of the first preference evaluation value is utilized and the final preference evaluation value can be calculated only with the first preference evaluation value.

7(b) is a diagram illustrating a case where the manipulation proficiency is less than the predetermined proficiency. Referring to FIG. 7(b), the predetermined reference value X10 of the amount of operation information collected may be half of the predetermined collection amount.

However, it is not limited thereto, and the predetermined reference value can be variously modified at a level obvious to those skilled in the art.

When the amount of motion information collected is less than the predetermined reference value X10, the first preference evaluation value and the second preference evaluation value are adjusted according to the first ratio B11 to calculate the final preference evaluation value. However, when the amount of motion information collected is greater than or equal to the predetermined reference value, the first preference evaluation value and the second preference evaluation value may be adjusted according to the third ratio B13 to calculate the final preference evaluation value.

For this reason, for example, even if 100% of the motion information is collected in advance, 75% of the first preference evaluation value and 25% of the second preference evaluation value are utilized to calculate the final preference evaluation value.

In this way, a more accurate preference evaluation value can be calculated by differentiating the use ratio of operation information and personal information according to the manager's proficiency in operating the container replica recommendation system.

The preference evaluation module can calculate preference evaluation values individually for each manager.

The preference evaluation module may change the above-described method of calculating the preference evaluation value according to whether the initial range and the correction range are selected.

If the number of pods within the initial range that are not included in the modified range more than a predetermined number of times after the initial and modified ranges are recommended are set as replicas by the manager, the preference evaluation module can change the method of calculating the preference evaluation value.

Here, the predetermined number of times may be 5 times. However, it is not limited to this, and the specific number of predetermined times can be variously modified at a level obvious to those skilled in the art.

As a specific example, if the number of Pods within the initial range that are not included in the modified range for more than a predetermined number of times after the initial and modified ranges are recommended are set as replicas by the manager, the preference evaluation module can designate as the preference evaluation value the average of past preference evaluation values of other managers who have the same age, gender, home region, and major as at least one or more of the same age, gender, hometown, and major as the manager when the predetermined collection amount is not collected.

This is a situation in which the reliability of collected personal information data is not trusted, and may be to calculate a preference evaluation value based on other reliable data.

Past preference evaluation values, which are preference evaluation values of other managers, may be matched with personal information of other managers and stored in the storage module.

As another example, if the number of pods within the initial range that are not included in the modified range for more than a predetermined number of times are set as replicas by the manager after the initial and modified ranges are recommended, the recommendation module can calculate the modified range by setting the average of the replica values selected by the administrator as the above-mentioned median value when operation information is collected as much as the predetermined collection amount.

That is, the recommendation module may calculate the correction range value by reflecting the preference evaluation value to the average value of values within the initial range or values not included in the correction range.

The recommendation module sets the average value of values within the initial range or values not included in the correction range as the median value, and adds the value obtained by multiplying the difference between the maximum value of the initial range and the median value by the efficiency ratio of the preferred evaluation value to the changed median value to recalculate the maximum value of the correction range.

In addition, the recommendation module sets the average value of the values within the initial range or values not included in the correction range as the median value, and subtracts the value obtained by multiplying the difference between the minimum value of the initial range and the median value by the stability weight of the preferred evaluation value to the changed median value. The minimum value of the correction range can be recalculated.

As a manager's disposition may vary depending on the task at hand, the extent of correction can be recalculated as part of the feedback.

In the accompanying drawings, in order to more clearly express the technical idea of the present invention, components that are not related to or detached from the technical idea of the present invention are briefly expressed or omitted.

In the above, the configuration and characteristics of the present invention have been described based on the embodiments according to the present invention, but the present invention is not limited thereto, and various changes or modifications can be made within the spirit and scope of the present invention.

110: collection module 120: analysis module
130: recommendation module 140: prediction module

Claims (10)

In the container replica recommendation system for operating and managing a cloud server providing container-based cloud services,
A collection module for collecting past use information, which is information about resources of a physical server used by a container;
an analysis module processing the past usage information by a predetermined processing method to calculate trend information, which is resource tendency;
a prediction module for calculating predicted usage information, which is information on resources of the physical server predicted to be used during a predetermined future period, based on the trend information and current usage information, which is information on resources of the currently operating physical server; and
A recommendation module that calculates an initial range, which is a range of the number of pods that can currently be operated on any worker node, based on the predicted usage information;
The recommendation module,
If the resources of the currently operating physical server are less than or equal to a predetermined ratio of the available resources of the physical server, the first initial range is recommended, and if the resources of the currently operating physical server exceed a predetermined ratio of the available resources of the physical server, the second initial range is recommended.
Container replica recommendation system.
According to claim 1,
The predetermined processing method,
A method of separating the past usage information into trend information, seasonality information obtained by analyzing a certain pattern according to the date, and irregular information that does not have a certain pattern,
Container replica recommendation system.
According to claim 1,
The recommendation module,
Calculating a second initial range by utilizing first allowable usage information calculated by adding or subtracting a first allowable value to the predicted usage information;
Calculating first initial information by utilizing second allowable usage information calculated by adding or subtracting a second allowable value to the predicted usage information;
Container replica recommendation system.
According to claim 3,
The first allowable value is,
smaller than the second allowable value,
Container replica recommendation system.
According to claim 1,
A control module for controlling worker nodes including containers so that the number of pods is maintained at the number set within the initial range;
The prediction module,
Calculating predicted usage information, which is information about resources of the physical server predicted to be used in a period after the predetermined future period;
The control module,
Increasing or decreasing the number of pods that can be run on any worker node according to the increase or decrease of the predicted usage information.
Container replica recommendation system.
In the container replica recommendation system for operating and managing a cloud server providing container-based cloud services,
A collection module for collecting past use information, which is information about resources of a physical server used by a container;
an analysis module processing the past usage information by a predetermined processing method to calculate trend information, which is resource tendency;
a recommendation module that calculates an initial range, that is, a range of the number of pods that can currently be operated on an arbitrary worker node, based on the trend information; and
A preference evaluation module for calculating a preference evaluation value related to management preferences of the manager of the cloud server by a predetermined preference calculation method;
The collection module,
Collect operation information, which is information generated while the physical server is operated by the manager,
The predetermined preference calculation method,
A method of calculating the preference evaluation value based on the operation information,
The recommendation module,
Calculating a correction range by correcting the initial range in consideration of the preference evaluation value,
Container replica recommendation system.
According to claim 6,
The preference evaluation value is,
Calculated in consideration of stability and efficiency,
Container replica recommendation system.
According to claim 6,
The collection module,
When the operation information is not collected as much as the predetermined collection amount, the personal information of the manager is collected,
The predetermined preference calculation method,
A method of calculating the preference evaluation value by considering the personal information and the operation information together,
Container replica recommendation system.
According to claim 8,
Further comprising: a proficiency judgment module for determining a manager's proficiency in operating the cloud server based on the operation information;
The predetermined preference calculation method,
A method of calculating the preference evaluation value by changing weights to which the personal information and the motion information are applied based on the manipulation proficiency,
Container replica recommendation system.
In the container replica recommendation method for operating and managing a cloud server implemented by a container replica recommendation system and providing a container-based cloud service,
collecting, by a collection module, past use information, which is information about resources of a physical server used by using a container;
Calculating, by an analysis module, trend information that is a resource tendency from the past usage information using a predetermined processing method;
Calculating predicted usage information, which is information on resources of the physical server predicted to be used during a predetermined future period, based on the trend information and current usage information, which is information on resources of the currently operating physical server, by a prediction module; and
A recommendation module calculates and recommends an initial range, which is a range of the number of Pods that can currently be operated on a worker node, based on the predicted usage information,
The recommendation module,
If the resources of the currently operating physical server are less than or equal to a predetermined ratio of the available resources of the physical server, the first initial range is recommended, and if the resources of the currently operating physical server exceed a predetermined ratio of the available resources of the physical server, the second initial range is recommended.
How to recommend a container replica.

Images