WO2019117593A1

WO2019117593A1 - Cloud computing device and method for supporting container-based resource allocation

Info

Publication number: WO2019117593A1
Application number: PCT/KR2018/015700
Authority: WO
Inventors: 김진용
Original assignee: 주식회사 넥스클라우드
Priority date: 2017-12-13
Filing date: 2018-12-11
Publication date: 2019-06-20
Also published as: KR20190070659A

Abstract

The present invention relates to a cloud computing device for supporting container-based resource allocation, comprising: a host unit for managing at least one container respectively performing tasks; a resource scheduling unit, which monitors resources related to a processing element and a memory element, that are used in the host unit, so as to detect an abnormality related to the monitored resources on the basis of a resource model studied in advance, and predicts resources necessary for the host unit when the abnormality is detected; a cluster management unit for reallocating, to the host, resources related to the processing element and the memory element; and a container orchestration unit, which determines actual resources necessary for the host on the basis of the predicted resources so as to request the cluster management unit to reallocate the resources.

Description

Cloud computing device and method supporting container-based resource allocation

The present invention relates to a cloud computing technology, and more particularly, to a cloud computing device and method for supporting container-based resource allocation that can improve the efficiency of individual resource use and improve cloud service efficiency.

Container cluster management technology is advantageous in terms of overhead, portability and efficiency based on container as compared to server virtualization technology. As a result, the transition to a container-based cloud environment is proceeding rapidly.

Conventional container cluster management technology can increase the infrastructure efficiency by dynamic allocation of hosts through the introduction of container-based private cloud, but by performing static allocation determined by the initial distribution of the resources of the container itself, And there is a disadvantage that only passive processing of a fault can be performed. This conventional technology has a difficulty in grasping the inefficient use of over-allocated resources, and thus has a limitation in lowering the service efficiency provided through the cloud service.

Korean Patent Laid-Open Publication No. 2017-0078012 is related to a method and system for optimizing resource allocation through monitoring and predicting cloud resources, including a cloud unit for providing cloud services as instances configured by virtualizing hardware resources, And a resource allocation optimization unit for predicting resource usage based on the resource usage monitored by the resource monitoring unit and the resource monitoring unit for monitoring and changing the hardware resource allocation for the instances based on the predicted resource usage.

Korean Patent No. 10-1781063 relates to a second-stage resource management method and apparatus for dynamic resource management, which receives a task to be executed in a cluster system, generates a first idle task corresponding to the input task, Generating a first virtual machine corresponding to the generated first idle job, executing the input job in the first virtual machine created, and executing the input job, the virtual node of the first virtual machine So that the virtual node can be adjusted dynamically according to the use state of the resource when the task is executed, thereby reducing the resource utilization efficiency of the system And the performance degradation due to resource shortage can be prevented.

[Prior Art Literature]

[Patent Literature]

Korean Patent Publication No. 2017-0078012

Korean Patent No. 10-1781063

An embodiment of the present invention is to provide a cloud computing apparatus and method supporting container-based resource allocation that can improve the efficiency of individual resource use and improve cloud service efficiency.

An embodiment of the present invention is to provide a cloud computing apparatus and method supporting container-based resource allocation that can improve prediction accuracy of resource use based on machine learning performed in connection with an abnormal symptom.

Among the embodiments, a cloud computing device supporting container-based resource allocation may include a host unit that manages at least one container, each of which performs tasks, a resource associated with processing elements and memory elements used in the host unit, A resource scheduling unit for detecting an abnormal symptom of the monitored resource based on the learned resource model and for predicting a resource required for the host unit when the abnormal symptom is detected; and a resource scheduling unit for informing the host unit about the processing element and the memory element And a container orchestration unit for determining a real resource required for the host unit based on the predicted resource and requesting the cluster management unit for the resource reallocation.

The resource scheduling unit may analyze the resource usage used by each of the at least one container when the abnormal symptom is detected.

The resource scheduling unit may apply the resource model to each of the at least one container to detect a specific container most highly associated with the abnormal symptom.

The resource scheduling unit may update the resource model based on a resource state collected for a predetermined period before the detection of the resource used by the specific container to predict a resource state at a specific time in the future.

The resource scheduling unit may update the resource model before the update based on the actual resource situation when the predicted resource situation has an error greater than or equal to the actual resource condition at the specific time point.

The resource scheduling unit may dynamically calculate the amount of the required resource by tracking usage increase / decrease of each of a plurality of resources used by the specific container while the abnormal symptom is continued.

The resource scheduling unit may update the resource model on the basis of a resource condition collected during the duration of the abnormal symptom when the abnormal symptom is solved before the specific time through the dynamic calculation of the predicted amount.

The resource scheduling unit may perform a pattern analysis on the increase / decrease of the usage amount during a specific time interval to predict whether there is an emergency shortage resource.

The container orchestration unit may track the actual resource status of the processing element and the memory element through the cluster management unit and may request the resource reallocation by selecting at least some of the predicted resources based on the actual resource status have.

Among the embodiments, a cloud computing method supporting container-based resource allocation is performed by a cloud computing device supporting container-based resource allocation, and the cloud computing method supporting container-based resource allocation includes (a) Managing at least one container performing the task; (b) monitoring resources associated with the processing elements and memory elements used in the host unit; and (C) performing a resource reallocation to the host with respect to the processing element and the memory element; and (d) comparing the predicted resource with the expected resource Based on the actual resources required for the host And requesting the cluster management unit to reallocate the resource.

The disclosed technique may have the following effects. It should be understood, however, that the scope of the disclosed technology is not to be construed as limited thereby, as it does not imply that a particular embodiment should include all of the following effects or merely include the following effects.

The cloud computing apparatus and method supporting container-based resource allocation according to an embodiment of the present invention can improve efficiency of individual resource use and improve cloud service efficiency.

The cloud computing apparatus and method supporting container-based resource allocation according to an embodiment of the present invention can improve the prediction accuracy of resource use based on machine learning performed in connection with an abnormal symptom.

FIG. 1 is a diagram illustrating a cloud computing apparatus supporting container-based resource allocation according to an embodiment of the present invention. Referring to FIG.

2 is a flowchart illustrating a process of performing resource allocation management by a cloud computing device supporting container-based resource allocation shown in FIG.

3 is an exemplary structure diagram of a cloud computing device supporting container-based resource allocation.

4 is a flowchart of a resource allocation process performed by a cloud computing device supporting container-based resource allocation.

5 is a diagram for explaining a process of dynamic allocation of container resources according to a resource allocation process.

A host unit which manages at least one container, each of which carries out the task;

Monitoring resources related to the processing elements and the memory elements used in the host unit, detecting an abnormal symptom of the monitored resource based on the learned resource model, and estimating resources required for the host unit when the abnormal symptom is detected A resource scheduling unit;

A cluster management unit for executing a resource reallocation to the host unit regarding the processing element and the memory element; And

And a container orchestration unit for determining an actual resource required for the host unit based on the predicted resource and requesting the cluster management unit for the resource reallocation.

The description of the present invention is merely an example for structural or functional explanation, and the scope of the present invention should not be construed as being limited by the embodiments described in the text. That is, the embodiments are to be construed as being variously embodied and having various forms, so that the scope of the present invention should be understood to include equivalents capable of realizing technical ideas. Also, the purpose or effect of the present invention should not be construed as limiting the scope of the present invention, since it does not mean that a specific embodiment should include all or only such effects.

Meanwhile, the meaning of the terms described in the present application should be understood as follows.

The terms "first "," second ", and the like are intended to distinguish one element from another, and the scope of the right should not be limited by these terms. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

It is to be understood that when an element is referred to as being "connected" to another element, it may be directly connected to the other element, but there may be other elements in between. On the other hand, when an element is referred to as being "directly connected" to another element, it should be understood that there are no other elements in between. On the other hand, other expressions that describe the relationship between components, such as "between" and "between" or "neighboring to" and "directly adjacent to" should be interpreted as well.

It is to be understood that the singular " include " or "have" are to be construed as including the stated feature, number, step, operation, It is to be understood that the combination is intended to specify that it does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

In each step, the identification code (e.g., a, b, c, etc.) is used for convenience of explanation, the identification code does not describe the order of each step, Unless otherwise stated, it may occur differently from the stated order. That is, each step may occur in the same order as described, may be performed substantially concurrently, or may be performed in reverse order.

The present invention can be embodied as computer-readable code on a computer-readable recording medium, and the computer-readable recording medium includes all kinds of recording devices for storing data that can be read by a computer system . Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like. In addition, the computer-readable recording medium may be distributed over network-connected computer systems so that computer readable codes can be stored and executed in a distributed manner.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Commonly used predefined terms should be interpreted to be consistent with the meanings in the context of the related art and can not be interpreted as having ideal or overly formal meaning unless explicitly defined in the present application.

Referring to FIG. 1, a cloud computing device 100 supporting container-based resource allocation includes a host unit 110, a resource scheduling unit 120, a cluster management unit 130, a container orchestration unit 140, (150).

The host unit 110 manages at least one container 112, each of which performs a task. In one embodiment, the host unit 110 includes a processing element 152 and a memory element 154, each of which is allocated as an allocation resource from the resource unit 150 under the control of the cluster management unit 130 and the container orchestration unit 140. [ The container managing unit 130 may create at least one container 112 that performs a task using at least a part of the container management unit 130 and constructs a container environment based on a doker to generate resource scheduling information based on the resource scheduling information received from the cluster managing unit 130 May sequentially perform each of the at least one task.

In one embodiment, the host unit 110 is capable of collecting information about at least some of the resources used through each of the at least one containers 112, the applications launched through each of them, and the tasks performed through the applications And may manage the collection agent 114. In one embodiment, the collection agent 114 may collect and provide information about the resource usage and performance of each of the at least one container 112 executing in the host unit 110, for example, cAdvisor (Container Advisor), MetricBeat, and FileBeat.

The resource scheduling unit 120 monitors the resources associated with the processing element 152 and the memory element 154 used in the host unit 110 to detect anomalies related to the monitored resource based on the resource model learned .

In one embodiment, the resource scheduling unit 120 receives and monitors resource usage and performance information regarding the container 112 running on the host unit 110 from the collection agent 114 of the host unit 110 And it is possible to detect an abnormal symptom related to the monitored resource by applying information on the received resource usage and performance to the learned resource model. For example, the resource scheduling unit 120 may detect, as an abnormal symptom, whether or not to use an over-allocated specific allocated resource inefficiently based on the resource usage collected from the collection agent 114. [

In one embodiment, the resource scheduling unit 120 may receive and monitor information about the overall resource usage allocated to the host unit 110 from the container orchestration unit 140, for example, As an abnormal symptom, at least one of the total resources in the specific host unit 110 that is overly allocated or ineffective.

The resource scheduling unit 120 can detect the presence or absence of at least one of the allocated resources allocated to the host unit 110 based on the collected resource usage, that is, the deficient resource, the emergency deficient resource, the excess resource, and the surplus resource, Based on the resource model, it is possible to analyze the collected resource status for a certain period of time to determine the presence of an abnormal symptom. For example, the resource scheduling unit 120 may determine the occurrence of a current problem when a tributary resource being used by the specific host unit 110 is detected, and analyze the resource status collected before and after the determination by the corresponding host unit 110 The possibility of occurrence of a problem after a specific time can be checked to determine the presence or absence of an abnormal symptom.

The resource scheduling unit 120 may analyze the resource usage and performance information collected during a specific time interval as big data to learn a resource model for detecting an abnormal symptom. In one embodiment, the resource scheduling unit 120 may collect information about resource usage and performance used by each host unit 110 from the collection agent 114, and may include information about resource usage collected during a particular time period, Based on the information about the performance, it is possible to perform the big data analysis according to the pre-stored resource learning algorithm to update the previously stored resource model, and to perform the resource model update by periodically or non-periodically re-learning. In one embodiment, the resource scheduling unit 120 may perform an update of the resource model upon detection of anomalous indications, which may be described in more detail in the section on anomalous indications.

The resource scheduling unit 120 can learn resource usage prediction models for predicting future resource usage associated with detection of anomalous symptoms by analyzing information on resource usage and performance collected during a specific time period as big data, It is possible to update the previously stored resource use prediction model by performing big data analysis according to the use prediction learning algorithm.

The resource scheduling unit 120 may include at least one of a data collection module, a batch analysis processing module, a real time analysis processing module, and a data repository for the big data analysis. For example, the data acquisition module can be implemented through a data acquisition algorithm based on Kafaka / Akka, and the real-time analysis processing module can be implemented through an analysis processing algorithm based on Spark Streaming. The data repository can store resource usage used by each host unit 110 and resource usage used by each container 112, and can store container log data. For example, the influx DB, which stores time series data, A Prometheus of a pull method, an Elastic Search based on a column collection, and a Cassandra of a NoSQL method, which periodically fetches data by connecting to a client.

The resource scheduling unit 120 predicts resources required for the host unit 114 when an abnormal symptom is detected. More specifically, the resource scheduling unit 120 may predict the amount of resources required to resolve the anomaly indication for each of the at least one host unit 114 associated with the anomalous symptom when an anomalous indication is detected, and in one embodiment, The necessary resource scheduling information can be generated by predicting the required resource amount based on the learned resource use prediction model.

The resource scheduling unit 120 may analyze the resource usage used by each of the at least one container 112 when an anomaly is detected. In one embodiment, the resource scheduling unit 120 may apply a resource model for each of the at least one container 112 to detect a particular container most closely associated with the anomaly indication. For example, the resource scheduling unit 120 may be configured to monitor the usage and performance of the processing element 152 and the memory element 154 to determine if the resource has been degraded below the reference performance, If a symptom is detected, the resource usage of each of at least one container 112 that is in use and allocated to the resource may be analyzed to detect a specific container that is using the resource in excess of the second reference usage amount.

The resource scheduling unit 120 may update the resource model based on the resource state collected for a certain period of time before the detection of the resource used by the detected specific container to predict the resource state at a specific time in the future. For example, when an abnormal symptom is detected at the first time point, the resource scheduling unit 120 may detect the first container of the first host unit among all the host units 110 as one most related to the abnormal symptom, The current resource model is updated by reflecting the data collected from the collection agent 114 up to the first time point before a predetermined time interval before the first time point with respect to the resources used by the first container, The resource usage of the first container can be predicted. In other words, the resource scheduling unit 120 can update the resource model to collect the collected data of the specific container most related to the anomalous symptom, so as to more accurately predict a future abnormal symptom. Here, the specific time point and the predetermined time point can be set and adjusted by the designer or the user in the future.

The resource scheduling unit 120 can update the resource model before the update based on the actual resource situation when the predicted resource situation is different from the actual resource condition at the specific time and the error tolerance range. For example, the resource scheduling unit 120 may generate a second resource model based on the first resource model at a first point of time following the detection of an anomalous symptom, calculate a second resource model at a second point of time corresponding to a specific point in time after the first point in time, And if the error in the error tolerance range is detected by comparing and analyzing the predicted resource situation at the second time point and the actually observed resource situation at the second time point and restoring to the first resource model instead of the second resource model And the third resource model can be generated by updating the restored first resource model by reflecting the actually observed resource situation.

In one embodiment, the resource scheduling unit 120 may adjust the future specific time and period based on the frequency of occurrence of the error in the recent specific time period, for example, based on the following equation (1) It is possible to adjust the downward adjustment of the specific time point in the future so as to be relatively inversely proportional to the frequency of occurrence,

[Equation 1]

_{t = (f 0 / f)} * t 0

_{p = (f / f 0)} * p 0

T ₀ , f _0, and p ₀ denote a reference time point, a reference error occurrence frequency, and a reference time point set by a designer or an administrator, respectively, Reference period)

The resource scheduling unit 120 may track the usage increase and decrease of each of a plurality of resources used by the specific container while the abnormal symptom is continued to dynamically calculate the predicted amount of the required resource. In one embodiment, each time a particular resource exceeding a reference amount of usage or a reference increase rate is detected among a plurality of resources used by the particular container, the resource scheduling unit 120 may change the required amount of the resource to a usage amount Can be dynamically calculated based on the growth rate. The resource scheduling unit 120 discretely detects a usage ratio, a usage increase / decrease amount and an increase / decrease transition with respect to each of a plurality of allocated resources allocated to the specific container in accordance with a time series, Can be calculated flexibly.

The resource scheduling unit 120 can update the resource model based on the resource conditions collected during the duration of the abnormal symptom when the abnormal symptom is solved before the specific point of time through the dynamic calculation of the predicted amount. For example, the resource scheduling unit 120 can dynamically calculate the predicted amount of resources predicted at the second point in time according to the anomaly detected at the first point of time, and calculate the third point of time before the second point in time The resource model can be updated based on the resource status collected through the collection agent 114 of the host unit 110 between the first and third time points.

The resource scheduling unit 120 can predict the presence of an emergency shortage resource through a pattern analysis on the usage increase / decrease of each of a plurality of resources during a specific time interval. In one embodiment, If a satisfying resource is detected, the corresponding resource is determined to be an emergency shortage resource that can be generated within a specific time in the future, and the emergency resource shortage resource may be reallocated preferentially in the resource reallocation process by reflecting it in the prediction of the resource situation.

The resource scheduling unit 120 can predict the presence or absence of excess resources or surplus resources through a pattern analysis on usage increase / decrease of each of a plurality of resources during a specific time interval, and if so, The excess resources may be allocated in a reduced manner or the redundant resources may be further allocated.

The cluster management unit 130 performs a resource reallocation on the processing element 152 and the memory element 154 to the host unit 110. [ The cluster management unit 130 can receive the resource status from the host unit 110 and manage the host resource status of the resources allocated to each host unit 110 and execute the execution Upon receipt of the request, host allocation can be performed to allocate a specific host unit 110 for execution of the task based on the host resource status. In one embodiment, when the resource reallocation request regarding the specific host unit 110 already allocated from the container orchestration unit 140 is received, the cluster management unit 130 updates the host identifier in the received resource reallocation request, And perform resource reallocation to the host unit 110 based on the requested resource amount.

The container orchestration unit 140 determines the actual resources required for the host unit 110 based on the predicted resources and requests resource allocation to the cluster management unit 130. [ In one embodiment, the container orchestration unit 140 analyzes the resource status currently in use when a predicted resource is received with respect to a specific container 112 executed in the specific host unit 110, It is possible to determine an allocatable real resource and perform a resource rescheduling according to the determined real resource to request a resource reallocation.

The container orchestration unit 140 tracks the actual resource status of the processing element 152 and the memory element 154 through the cluster management unit 130 and selects at least a part of the corresponding resources based on the actual resource status To request resource reassignment. For example, the container orchestration unit 140 may determine that additional use of three CPUs and 50 GB of memory is predicted by the resource scheduling unit 120 for a specific host unit 110, If the current state is 2 CPUs and 30 GB of memory, resource reallocation can be requested by reflecting the actual resource status.

The resource unit 150 may include at least one processing element 152 and at least one memory element 154 as resources. Each of the at least one processing element 152 may be implemented as a processor capable of independently processing a task performed by the host unit 110 when assigned to the host unit 110. [ Each of the at least one memory element 154 may be implemented as a memory that can be created or stored throughout the task execution by the host unit 110 when the host unit 110 is allocated to the host unit 110. For example, A nonvolatile memory that can function as an auxiliary storage device, and a main storage device that can function as a main storage device.

Each of the host unit 110, the resource scheduling unit 120, the cluster management unit 130, the container orchestration unit 140 and the resource unit 150 may be implemented as independent computing devices or at least some of them may be functionally or physically integrated May be implemented as a computing device, and at least some of them may function as a server.

In Fig. 2, the host unit 110 manages at least one container 112, each of which performs a task (step S210).

The resource scheduling unit 120 monitors the resources associated with the processing element 152 and the memory element 154 used in the host unit 110 and detects anomalous indications of the monitored resources based on the learned resource model (Step S220).

The resource scheduling unit 120 predicts resources required for the host unit 114 when an abnormal symptom is detected (step S230).

The cluster management unit 130 performs a resource reallocation with respect to the processing element 152 and the memory element 154 to the host unit 110 (step S240).

The container orchestration unit 140 determines an actual resource required for the host unit 110 based on the predicted resource, and requests resource allocation to the cluster management unit 130 (step S250).

3 to 5 are views for explaining an embodiment of a cloud computing device supporting container-based resource allocation shown in FIG. 1 in more detail.

3 illustrates an exemplary implementation of a cloud computing device 100 that supports container-based resource allocation. FIG. 4 illustrates an exemplary implementation of a container- FIG. 5 is a diagram illustrating a process of performing dynamic allocation of container resources according to the resource allocation process. Referring to FIG.

In FIG. 3, the resource scheduling unit 120 may be implemented through the big data analysis module 120a and the resource allocation support module 120b. As described above, the resource allocation support module 120b is configured to (1) automatically allocate resources for prediction resource allocation, auto-scaling and performance verification, (2) monitor performance, alert anomalies, Dashboard function for advance power management (APM), and (3) machine learning for predicting required resources, detecting abnormal symptoms, and analyzing excess resources.

In FIG. 4, the resource scheduling unit 120 may collect resources and log data used through the collection agent 114 of the host unit 110 to perform resource monitoring on the resource status, To predict resource use, to analyze excess resources, or to detect anomalous indications. The resource scheduling unit 120 can notify the resource status through the dashboard, perform performance monitoring on the resource to be used, and provide a notification message about the occurrence of the problem through an alarm or SMS when a problem occurs . The resource scheduling unit 120 may send a warning alarm or request a reallocation if an emergency resource shortage is detected and may provide a notification message when an abnormal symptom is detected, You can update it periodically. When an abnormal symptom is detected, the resource scheduling unit 120 may determine at least one of a resource requirement amount for an emergency shortage resource, a resource requirement amount based on a resource use prediction model, a resource reduction amount for an excess resource, and a resource addition allocation amount for a surplus resource And provides the information to the container orchestration unit 140 as information for resource reassignment. The container orchestration unit 140 can request the cluster management unit 130 to allocate resources on the basis thereof and cause the cluster management unit 130 to perform the resource relocation.

In FIG. 5, the resource scheduling unit 120 may track the usage increase / decrease of a plurality of resources used by each of the at least one containers 112 of the host unit 110 to dynamically calculate a predicted amount of necessary resources, The container orchestration unit 140 can increase the resource allocation efficiency by making the resource reallocation of the predicted amount of the required resource dynamically performed. Accordingly, the cloud computing device 100 supporting the container-based resource allocation can improve the efficiency of the cloud service by grasping the inefficient use of the excessively allocated resources, and can improve the efficiency of the cloud service in a container- The efficiency of individual operating resources can be improved.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as set forth in the following claims And changes may be made without departing from the spirit and scope of the invention.

[Description of Symbols]

100: Cloud computing device supporting container-based resource allocation

110: Host unit 120: Resource scheduling unit

130: Cluster management unit 140: Container orchestration unit

150: resource unit

Claims

A host unit which manages at least one container, each of which carries out the task;

Monitoring resources related to the processing elements and the memory elements used in the host unit, detecting an abnormal symptom of the monitored resource based on the learned resource model, and estimating resources required for the host unit when the abnormal symptom is detected A resource scheduling unit;

A cluster management unit for executing a resource reallocation to the host unit regarding the processing element and the memory element; And

And a container orchestration unit for determining an actual resource required for the host unit based on the predicted resource and requesting the cluster management unit for the resource reallocation.
The apparatus of claim 1, wherein the resource scheduling unit

And analyzing resource usage used by each of the at least one container when the abnormal symptom is detected.
3. The apparatus of claim 2, wherein the resource scheduling unit

Wherein the resource model is applied to each of the at least one container to detect a specific container most related to the abnormal symptom.
4. The apparatus of claim 3, wherein the resource scheduling unit

Wherein the resource model updating unit estimates a resource state at a specific point in time by updating the resource model based on a resource state collected for a predetermined period before the detection of the resource used by the specific container.
5. The apparatus of claim 4, wherein the resource scheduling unit

And updates the resource model before the update based on the actual resource situation when the predicted resource situation is different from an actual resource situation at the specific time and an error over an error tolerance range.
4. The apparatus of claim 3, wherein the resource scheduling unit

Wherein the controller is configured to dynamically calculate a predicted amount of the required resources by tracking usage increase / decrease of each of a plurality of resources used by the specific container while the abnormal symptom continues.
7. The apparatus of claim 6, wherein the resource scheduling unit

Wherein the resource model updating unit updates the resource model based on the resource state collected during the duration of the abnormal symptom when the abnormal symptom is solved before the specific time through the dynamic calculation of the predicted amount.
7. The apparatus of claim 6, wherein the resource scheduling unit

Wherein a pattern analysis on the usage increase / decrease is performed during a specific time interval to predict the presence / absence of an emergency shortage resource.
The apparatus of claim 1, wherein the container orchestration unit

Wherein the cluster management unit tracks the actual resource status of the processing element and the memory element through the cluster management unit and requests the resource reallocation by selecting at least some of the predicted resources based on the actual resource status. Computing device.
A cloud computing method supporting container-based resource allocation performed by a cloud computing device supporting container-based resource allocation,

(a) managing at least one container each performing a task;

(b) monitoring resources associated with the processing elements and memory elements used in the host unit, detecting anomalous indications of the monitored resources based on the learned resource model, and Predicting resources;

(c) performing a resource reassignment for the processing element and the memory element to the host; And

(d) determining an actual resource required for the host based on the predicted resource, and requesting the resource management unit to allocate the resource to the cluster management unit.