KR102320317B1 - Method for selecting predict-based migration candidate and target on cloud edge - Google Patents

Abstract

A cloud edge management method and a cloud edge device are provided. According to this cloud edge management method, when workload data is received, the probability of failure of a pod is calculated using the learned predictive model, and migration can be performed for the pod according to the calculated probability of failure. As a result, it is possible to prevent possible risks in advance through the management of migration candidates, thereby continuously providing operational data immediacy and service stability in the cloud edge environment.

Description

{Method for selecting predict-based migration candidate and target on cloud edge}

The present invention relates to a cloud edge management method and a cloud edge device, and more particularly, to a cloud edge management method and a cloud edge device for performing prediction-based migration in a cloud edge environment.

The content described in this section merely provides background information for the present embodiment and does not constitute the prior art.

Recently, cloud technology is being applied to IT services in various fields. However, when the cloud data center is physically far away, there is a problem in that the speed is slowed down and processing time is delayed due to a bottleneck.

In order to compensate for this, more powerful cloud edge technology is being provided by adding edge computing technology to cloud technology. Edge computing, also known as fog computing, is a method in which the edge (or fog, small cloud node) located at the edge of the network or near the user performs the main data analysis processing function and transmits only the processed results to the main cloud. .

However, in the cloud edge environment, resource availability support is not possible, and overload control according to cloud edge characteristics can only be processed through post-processing through message log analysis.

In addition, although resource analysis was in progress in a large-scale cloud environment, it is common that resource analysis is not performed in an edge environment.

In addition, most of the workload management and prediction in the cloud environment is for a large-scale cloud environment, and it is common that workload management and prediction for the edge environment are not made.

Therefore, the search for a method for workload management and prediction in such a cloud edge environment is requested.

The present invention has been devised to solve the above problems, and an object of the present invention is to calculate the probability of occurrence of a pod failure using a learned predictive model when the workload data is received, and the calculated failure It is to provide a cloud edge management method and cloud edge device that perform migration for the corresponding pod according to the possibility of occurrence.

The technical problems to be achieved in the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. will be able

In a cloud edge management method by a cloud edge management device according to an embodiment of the present invention for achieving the above object, when workload data is received, the probability of occurrence of a pod failure using the learned predictive model calculating; and performing migration on the corresponding pod according to the calculated failure probability.

Also, in the migration step, when the calculated failure probability is greater than or equal to the first threshold, data stored in the corresponding pod may be migrated to a virtual machine having more virtual resources.

In addition, in the migration step, when the calculated failure probability is less than or equal to the second threshold, data stored in the corresponding pod may be migrated to a virtual machine with fewer virtual resources.

And, using the workload data to learn the predictive model for predicting the possibility of failure through deep learning; may further include.

In addition, the learning may include learning the predictive model using a Long Short Term Memory (LSTM) algorithm.

In addition, the workload data may include workload data of nodes and pods, workload data of virtual machines, workload data of containers, and workload data of applications.

In addition, the workload data may include CPU usage data and memory usage data.

On the other hand, according to an embodiment of the present invention, a cloud edge management apparatus, a communication unit for receiving the workload data of the pod (Pod); and a control unit that, when the workload data is received, calculates the probability of occurrence of failure of the pod using the learned predictive model, and performs migration for the corresponding pod according to the calculated probability of occurrence of failure.

According to various embodiments of the present invention, when the workload data is received, the probability of occurrence of a pod is calculated using the learned predictive model, and the cloud performs migration for the corresponding pod according to the calculated probability of occurrence of failure. By being able to provide edge management methods and cloud edge devices, it is possible to prevent possible risks in advance through migration candidate management, thereby continuously providing operational data immediacy and service stability in the cloud edge environment.

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those of ordinary skill in the art from the following description. will be.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included as a part of the detailed description to help the understanding of the present invention, provide embodiments of the present invention, and together with the detailed description, explain the technical features of the present invention.
1 is a diagram illustrating a cloud edge environment, according to an embodiment of the present invention;
2 is a diagram illustrating a configuration of a cloud edge according to an embodiment of the present invention;
3 is a view showing the structure of a cloud edge management apparatus according to an embodiment of the present invention;
4 is a flowchart provided to explain a cloud edge management method according to an embodiment of the present invention;
5 is a diagram schematically illustrating a process in which a cloud edge management method is performed according to an embodiment of the present invention;
6 is a diagram illustrating an example of predicting the probability of failure for each pod in a cloud edge environment, according to an embodiment of the present invention.

In order to clarify the characteristics and advantages of the problem solving means of the present invention, the present invention will be described in more detail with reference to specific embodiments of the present invention shown in the accompanying drawings.

However, detailed descriptions of well-known functions or configurations that may obscure the gist of the present invention in the following description and accompanying drawings will be omitted. Also, it should be noted that throughout the drawings, the same components are denoted by the same reference numerals as much as possible.

The terms or words used in the following description and drawings should not be construed as being limited to conventional or dictionary meanings, and the inventor may appropriately define the concept of terms for describing his invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention based on the principle that there is.

Therefore, the embodiments described in this specification and the configurations shown in the drawings are only the most preferred embodiment of the present invention, and do not represent all the technical spirit of the present invention, so at the time of the present application, various It should be understood that there may be equivalents and variations.

In addition, terms including ordinal numbers such as first, second, etc. are used to describe various components, and are used only for the purpose of distinguishing one component from other components, and to limit the components. not used For example, without departing from the scope of the present invention, the second component may be referred to as the first component, and similarly, the first component may also be referred to as the second component.

In addition, when an element is referred to as being “connected” or “connected” to another element, it means that it is logically or physically connected or can be connected.

In other words, it should be understood that a component may be directly connected or connected to another component, but another component may exist in the middle, and may be indirectly connected or connected.

In addition, terms such as "comprises" or "have" described in this specification are intended to designate the existence of a feature, number, step, operation, component, part, or combination thereof described in the specification, one or the It should be understood that the above does not preclude the possibility of the existence or addition of other features or numbers, steps, operations, components, parts, or combinations thereof.

In addition, terms such as “…unit”, “…group”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. have.

Also, "a or an", "one", "the" and similar terms are otherwise indicated herein in the context of describing the invention (especially in the context of the following claims). or may be used in a sense including both the singular and the plural unless clearly contradicted by context.

Hereinafter, the present invention will be described in more detail with reference to the drawings.

1 is a diagram illustrating a cloud edge environment according to an embodiment of the present invention.

As shown in FIG. 1 , the cloud edge 10 is managed by the cloud edge management device 100 . Specifically, the cloud edge management device 100 manages the resource usage status, environment data, workload data, etc. of the cloud edge 10, and accordingly manages the migration of data of the pod of the cloud edge. will perform

The cloud edge management device 100 may be implemented not only as a physically independent device by itself, but may also be implemented in a form included as a part of any device or system, and a program or frame installed on a smartphone, computer, server, etc. Of course, it may be implemented in the form of software such as a work or an application. In addition, each component of the cloud edge management apparatus 100 may be implemented as a physical component or may be implemented as a component in the form of a function of software.

Cloud Edge is a technology that provides a more powerful cloud by adding edge computing technology to cloud technology. Edge computing, also known as fog computing, is a method in which the edges (or fog, nodes) located at the edge of the network or near the user perform main data analysis and processing functions, and only the processed results are transmitted to the main cloud.

As shown in FIG. 1 , it can be confirmed that the cloud edge 10 includes a plurality of nodes and includes at least one pod in the node. Hereinafter, the structure of the cloud edge will be described in more detail with reference to FIG. 2 .

In the cloud edge 10 , the edge is a node (node, server, etc.) located in the network closest to the user in the IT infrastructure. That is, the cloud edge 10 configures the infrastructure to transmit and process the data or content requested by the user closest to the user. For example, the content delivery network is a network service that utilizes the cloud edge concept, and most global content services including YouTube and Netflix provide content through the content delivery network.

Edge computing of the cloud edge 10 refers to a computing method that processes data with small facilities at multiple points, not a cloud computing method that centrally processes data. As we enter the era of the Internet of Things, the amount of data collected centrally through various routes increases, and in the Internet of Things environment where real-time processing is important, concentrated data must be processed without delay. It is impossible to accept In order to compensate for this problem, the cloud edge 10 processes data directly from 'nodes', which are edge servers installed in various places, and informs the main cloud of the results.

The cloud edge 10 has a shorter latency than the conventional cloud computing. Since data is processed directly from the nearest terminal or Internet of Things device, it can respond to situations without delay, and it can be used in the autonomous driving industry because it guarantees a fast response speed.

In addition, the cloud edge 10 can alleviate the security problem of the cloud to some extent. In the existing cloud, when a problem occurs in the central data center, all web/mobile services connected to it are affected. However, since the cloud edge 10 processes most of the calculations in each device, it can be considered safe in that attacking one system does not damage the entire service.

2 is a diagram illustrating a configuration of a cloud edge 10 according to an embodiment of the present invention. As shown in FIG. 2 , the cloud edge 10 includes a plurality of nodes 200 , and at least one pod 210 is included in the node.

The cloud edge 10 shown in FIG. 2 may be implemented using Kubernetes. In this case, the cloud edge management device 100 performs the same function as the master managing the entire cluster. All commands call the API server of the cloud edge management device 100 that is the master, and the node 200 performs necessary tasks while communicating with the cloud edge management device 100 . When commanding a container of a specific node 200 or inquiring a log, instead of giving a command to the node 200 directly, the cloud edge management device 100 gives a command and the cloud edge management device 100 accesses the node 200 . Instead, it responds with a result.

The node 200 is composed of one server or a small cloud composed of a plurality of servers. The node 200 creates a necessary pod 210 while communicating with the cloud edge management device 100 and configures a network and storage.

The pod 210 is where actual containers are created and can be expanded to hundreds or thousands. The pod 210 may define a purpose of use (GPU-specific, SSD server, etc.) by attaching a label to each. A pod 210 is the smallest unit that can be deployed in Kubernetes and has one or more container 211 , storage 213 , and network 215 properties. At least one container 211 belonging to the pod 210 may share the storage 213 and the network 215 and may access each other through a localhost.

The cloud edge 10 includes a plurality of nodes and pods having such a structure.

Hereinafter, the configuration of the cloud edge management apparatus 100 will be described in more detail with reference to FIG. 3 . 3 is a diagram illustrating the structure of a cloud edge management apparatus 100 according to an embodiment of the present invention.

As shown in FIG. 3 , the cloud edge management apparatus 100 includes an input unit 110 and a control unit 120 .

The communication unit 110 is communicatively connected to the cloud edge 10 , and workload data of a pod of the cloud edge 10 is received. Here, the workload data represents data on the amount of virtual resources used in the cloud edge 10 , and may include CPU usage data, memory usage data, storage usage data, network load data, and the like. The workload data may include workload data of nodes and pods, workload data of virtual machines, workload data of containers, and workload data of applications.

The communication unit 110 may perform communication in various wireless communication methods such as Bluetooth, Wi-Fi (WIFI), short-range wireless communication (NFC), cellular, and LTE (Long-Term Evolution). Of course it could be.

The controller 120 controls the overall operation of the cloud edge management apparatus 100 . Specifically, when the workload data is received, the controller 120 calculates the failure probability of the pod using the learned predictive model, and performs migration for the corresponding pod according to the calculated failure probability. Here, the failure probability is a value indicating the probability that a failure occurs in servers or equipment included in the data center, and the higher the failure probability value, the higher the probability of failure occurring within a certain period of time.

In addition, detailed operations of the control unit 120 will be described with reference to FIG. 4 . 4 is a flowchart provided to explain a cloud edge management method according to an embodiment of the present invention.

First, the controller 120 learns a predictive model for predicting the probability of failure for each pod through deep learning using workload data for each pod.

In this case, the controller 120 may train the predictive model using various deep learning algorithms, for example, the controller 120 may train the predictive model using a Long Short Term Memory (LSTM) algorithm. . Also, the controller 120 may train the predictive model by applying the Inference Pipeline, Policy, and Training Pipeline processes. For example, the control unit 120 inputs a workload data set when a failure does not occur for a certain period and a workload data set when a failure occurs during a certain period into the predictive model, and uses a deep learning algorithm to generate the predictive model. will learn

And, when the learning is completed, the controller 120 uses the learned predictive model. Specifically, when the workload data is received, the controller 120 uses the learned predictive model to cause a failure of the pod. The probability of occurrence is calculated for each pod (S420). The predictive model is an artificial intelligence deep learning learning model that calculates and outputs the probability of failure for each pod when workload data is input for each pod.

Thereafter, the controller 120 determines whether to perform migration for the corresponding pod according to the calculated failure probability, and performs migration when a preset condition is satisfied (S430). Specifically, when the calculated failure probability is greater than or equal to the first threshold, the controller 120 may migrate data stored in the corresponding pod to a virtual machine having more virtual resources. Also, when the calculated failure probability is less than or equal to the second threshold, the controller 120 may migrate data stored in the corresponding pod to a virtual machine having fewer virtual resources.

For example, the control unit 120 may migrate data stored in the corresponding pod to a virtual machine having more than twice as many virtual resources (CPU, memory, storage, etc.) can In addition, the controller 120 migrates the data stored in the pod to a virtual machine with less than 1/2 times less virtual resources (CPU, memory, storage, etc.) when the failure probability is less than or equal to the second threshold of 20%. can do. The first threshold value and the second threshold value can be set by a user according to a policy and are changeable values. In addition, the controller 120 may migrate the pods of the cloud edge 10 according to the possibility of failure in various ways.

Through this process, the cloud edge management device 100 learns and predicts the possibility of failure and migrates the data of the pod of the cloud edge, so that it is possible to minimize the failure management risk even in the cloud edge environment, and through migration candidate management By being able to prevent possible risks in advance, it will be possible to continuously provide the immediacy of operational data and service stability in the cloud edge environment.

5 is a diagram schematically illustrating a process in which a cloud edge management method is performed according to an embodiment of the present invention.

5 , the workload data 510 may include workload data of nodes and pods, workload data of virtual machines, workload data of containers, and workload data of applications. The workload data of a node and a pod indicates the CPU usage, memory usage, storage usage, network usage, etc. being used by the node corresponding to the corresponding pod and the corresponding pod, respectively. Container workload data indicates CPU usage, memory usage, storage usage, network usage, etc. being used by each of the containers included in the pod. The workload data of the virtual machine indicates CPU usage, memory usage, storage usage, network usage, etc. being used by each of the virtual machines included in the pod. The workload data of the application indicates the CPU usage, memory usage, storage usage, network usage, etc. each application included in the pod is using.

Such workload data 510 is input to the cloud edge device 100 . Then, the cloud edge device 100 may learn the predictive model by using a Long Short Term Memory (LSTM) algorithm, and specifically, may train the predictive model by applying the Inference Pipeline, Policy, and Training Pipeline processes.

Also, the cloud edge device 100 may display the probability of occurrence of a failure with gauges 520 and 530 for each pod.

Through this, the cloud edge device 100 can more intuitively display the possibility of occurrence of a failure to the user.

6 is a diagram illustrating an example of predicting the probability of failure for each pod in a cloud edge environment, according to an embodiment of the present invention.

6 shows an environment of the cloud edge 10 including two nodes (Node1 and Node2). And, node 1 (Node1) includes four pods (Pod11, Pod12, Pod13, Pod14), and node 2 (Node2) includes four pods (Pod21, Pod22, Pod23).

And, the probability of failure of Pod11 is 90%, the probability of failure of Pod12 is 55%, the probability of failure of Pod13 is 10%, the probability of failure of Pod14 is 75%, the probability of failure of Pod21 is 50%, and the probability of failure of Pod22 is 50%. It can be seen that the probability of occurrence is 80%, and the probability of failure of Pod23 is 15%. That is, if the first threshold is 80% and the second threshold is 20%, Pod11 and Pod22 are targets to be migrated to larger virtual machines, and Pod13 and Pod23 are migrated to smaller virtual machines. become a target for performing

Accordingly, the cloud edge device 100 migrates Pod11 and Pod22 to a virtual machine with more virtual resources, and Pod13 and Pod23 migrates to a virtual machine with fewer virtual resources.

Through this, the cloud edge device 100 is a virtual machine with more virtual resources for a pod that is likely to fail due to insufficient resources, and a virtual machine with fewer virtual resources for a pod that is less likely to fail due to sufficient resources. By migrating to a new location, it is possible to efficiently use the virtual resources of the cloud edge and minimize the possibility of failure based on prediction.

On the other hand, it goes without saying that the technical idea of the present invention can also be applied to a computer-readable recording medium containing a computer program for performing the function and method of the device according to the present embodiment. In addition, the technical ideas according to various embodiments of the present invention may be implemented in the form of computer-readable programming language codes recorded on a computer-readable recording medium. The computer-readable recording medium may be any data storage device readable by the computer and capable of storing data. For example, the computer-readable recording medium may be a ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical disk, hard disk drive, flash memory, solid state disk (SSD), or the like. In addition, the computer-readable code or program stored in the computer-readable recording medium may be transmitted through a network connected between computers.

Although this specification and drawings describe exemplary device configurations, implementations of the functional operations and subject matter described herein may be implemented in other types of digital electronic circuits, or include structures disclosed herein and structural equivalents thereof. may be implemented as computer software, firmware, or hardware, or a combination of one or more of these.

Accordingly, although the present invention has been described in detail with reference to the above-described examples, those skilled in the art to which the present invention pertains can make modifications, changes and modifications to the examples without departing from the scope of the present invention.

In addition, although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and the technical field to which the present invention belongs without departing from the gist of the present invention as claimed in the claims Various modifications are possible by those of ordinary skill in the art, and these modifications should not be individually understood from the technical spirit or prospect of the present invention.

10 : Cloud Edge
100: cloud edge management device
110: communication department
120: control unit

Claims (8)

In the cloud edge management method by the cloud edge management device,
When the workload data is received, calculating the probability of occurrence of a failure of the pod using the learned predictive model; and
Including; according to the calculated failure probability, performing migration for the pod;
The steps to perform the migration are:
When the calculated probability of failure is greater than or equal to the first threshold, the data stored in the pod is migrated to a virtual machine with more virtual resources. Migrate to fewer virtual machines,
A cloud edge management method, characterized in that the calculated failure probability is displayed as a gauge for each pod.
delete delete The method according to claim 1,
Further comprising; training a predictive model that predicts the possibility of failure using workload data through deep learning;
The calculation step is
A cloud edge management method, characterized in that the probability of occurrence of a pod failure is calculated using the predictive model learned through the deep learning.
5. The method according to claim 4,
The learning steps are
A cloud edge management method, characterized in that the predictive model is trained using an LSTM (Long Short Term Memory) algorithm.
The method according to claim 1,
workload data,
A cloud edge management method, comprising: workload data of nodes and pods, workload data of virtual machines, workload data of containers, and workload data of applications.
7. The method of claim 6,
workload data,
Cloud edge management method comprising CPU usage data and memory usage data.
Communication unit for receiving the workload data of the pod (Pod); and
When the workload data is received, a control unit that calculates the probability of occurrence of failure of the pod using the learned predictive model, and performs migration for the corresponding pod according to the calculated probability of occurrence of failure;
the control unit,
When the calculated probability of failure is greater than or equal to the first threshold, the data stored in the pod is migrated to a virtual machine with more virtual resources. Migrate to fewer virtual machines,
A cloud edge management device, characterized in that the calculated failure probability is displayed as a gauge for each pod.

Images