KR20240067674A - Container-based dynamic workload processing system, apparatus and method considering data local information - Google Patents

Abstract

A dynamic workload processing device according to an embodiment of the present invention includes a memory storing at least one command; and a processor that performs at least one instruction. At this time, the processor obtains task information necessary to support the user service request by executing at least one command, obtains local information of a data source for processing the task corresponding to the task information, and performs the task based on the local information. Dynamically processes the workload to process tasks corresponding to information.

Description

Container-based dynamic workload processing system, apparatus, and method considering data local information {CONTAINER-BASED DYNAMIC WORKLOAD PROCESSING SYSTEM, APPARATUS AND METHOD CONSIDERING DATA LOCAL INFORMATION}

The present invention relates to edge computing and micro services, and to container-based dynamic workload processing technology that enables big data processing considering data location.

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

According to the prior art, most applications were developed in a monolithic structure. Monolithic structure is a method of developing an application program as one large block, connecting various processes and databases required for the operation of the program, and organizing the program hierarchically. The monolithic structural style is closely linked to each internal process and is typically developed using a single design program.

On the other hand, when developing an application, Microservice Architecture divides it into multiple small services and merges the small services to complete one application. In other words, the microservice structure creates one application service by specializing each service by module and connecting and integrating each module using an interface.

When transmitting data collected from an Internet of Things (IoT) device equipped with a sensor to the cloud or server, existing devices transmit data by implementing a program in the form of firmware or an embedded OS (Operating System). Create software and transmit data using .

With the advent of the Internet of Things (IoT), significant amounts of data are generated in edge networks. The vast and diverse data, big data domain, has developed rapidly, and problems are arising in various areas, from complex business logic to the infrastructure required to process the ever-increasing volume, speed, and variety of data.

Edge computing technology aims to complement cloud computing by utilizing the hardware resources of edge devices. However, since solutions that process vast and diverse data (big data processing solutions) have traditionally been designed for cloud service workloads, their application to microservices and edge computing has been pointed out as unsuitable or inefficient.

Korean Patent Publication No. KR 10-2021-0060203 “Microservice structure reconstruction method and device” (publication date May 26, 2021)

The purpose of the present invention is to propose a container-based dynamic workload processing system, device, and method that can efficiently provide data services by considering location/region information of the data source.

The purpose of the present invention is to propose a cluster-based dynamic workload processing system, device, and method for edge server nodes that can efficiently provide data services by considering location/region information of data sources.

A dynamic workload processing device that dynamically processes a workload for providing a service according to an embodiment of the present invention for achieving the above object includes: a memory storing at least one command; and a processor that performs at least one instruction. At this time, the processor obtains task information necessary to support the user service request by executing at least one command, obtains local information of a data source for processing the task corresponding to the task information, and performs the task based on the local information. Dynamically processes the workload to process tasks corresponding to information.

When dynamically processing a workload, the processor may dynamically reconfigure the order of the workload and the scheduling of at least one of the workload performers based on local information.

In obtaining task information necessary to support a user service request, the processor may receive a task information request requesting task information necessary to support the user service request from a control layer in the first cluster, and request task information. By analyzing the user service request corresponding to , task information necessary to support the user service request can be obtained.

In obtaining local information of the data source, the processor may obtain local information of the data source for processing a task corresponding to task information from the data layer in the first cluster.

The processor may obtain information on a first proximate dynamic workload processing node located around the data source based on local information of the data source.

In dynamically processing the workload, the processor may dynamically process the workload based on local information of the data source and information of the first adjacent dynamic workload processing node.

The processor may obtain information of a second cluster including a second proximate dynamic workload processing node located around the data source based on local information of the data source.

When dynamically processing a workload, the processor may dynamically process the workload based on local information of the data source and information of the second cluster.

When there are a plurality of second proximate dynamic workload processing nodes, the processor may dynamically schedule the workload so that the plurality of second proximate dynamic workload processing nodes in the second cluster distribute and process tasks corresponding to the task information. there is.

A dynamic workload processing system according to an embodiment of the present invention includes a first dynamic workload processing node corresponding to a control layer in a first cluster; and a second dynamic workload processing node corresponding to a computational layer within the first cluster.

At this time, the first dynamic workload processing node receives the user service request, and the first dynamic workload processing node transmits a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node. And, the second dynamic workload processing node analyzes the user service request corresponding to the task information request to obtain task information necessary to support the user service request, and the second dynamic workload processing node performs the task corresponding to the task information. Local information of the data source for processing is acquired, and the second dynamic workload processing node dynamically processes the workload for processing the task corresponding to the task information based on the local information.

When the second dynamic workload processing node dynamically processes the workload, the order of the workload and the scheduling of at least one of the workload performers may be dynamically reconfigured based on local information.

The second dynamic workload processing node obtains local information of the data source, wherein the local information of the data source is used to process the task corresponding to the task information from the first proximate dynamic workload processing node corresponding to the data layer in the first cluster. Information can be obtained.

The second dynamic workload processing node may obtain information about a second adjacent dynamic workload processing node located around the data source based on local information of the data source.

When dynamically processing the workload, the second dynamic workload processing node may dynamically process the workload based on local information of the data source and information of the second adjacent dynamic workload processing node.

The second dynamic workload processing node may obtain information of a second cluster including a second adjacent dynamic workload processing node located around the data source based on local information of the data source.

The second dynamic workload processing node may dynamically process the workload based on local information of the data source and information of the second cluster.

The second dynamic workload processing node is configured to distribute the workload so that the plurality of second adjacent dynamic workload processing nodes in the second cluster distribute tasks corresponding to the task information when there is a plurality of second adjacent dynamic workload processing nodes. can be scheduled dynamically.

When there are a plurality of second dynamic workload processing nodes, one of the plurality of second dynamic workload processing nodes is selected based on at least one of the operation failure of the first dynamic workload processing node and the updated local information. 1 It can correspond to the control layer by replacing the dynamic workload processing node.

The dynamic workload processing method according to an embodiment of the present invention includes processing a user service request received via a first dynamic workload processing node corresponding to a control layer in a first cluster to a user service request corresponding to a computation layer in the first cluster. Analyzing by a second dynamic workload processing node to obtain task information necessary to support a user service request; Obtaining, by a second dynamic workload processing node, local information of a data source for processing a task corresponding to the task information; and a second dynamic workload processing node dynamically processing a workload for processing a task corresponding to task information based on local information.

In the step where the second dynamic workload processing node dynamically processes the workload, the order of the workload and the scheduling of at least one of the workload performers may be dynamically reconfigured based on local information.

A dynamic workload processing method according to an embodiment of the present invention includes the steps of: a first dynamic workload processing node receiving a user service request; It may further include transmitting, by the first dynamic workload processing node, a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node.

In the step of the second dynamic workload processing node acquiring local information of the data source, the data source for processing the task corresponding to the task information from the first proximate dynamic workload processing node corresponding to the data layer in the first cluster Local information can be obtained.

The dynamic workload processing method according to an embodiment of the present invention includes the second dynamic workload processing node obtaining information about a second adjacent dynamic workload processing node located around the data source based on local information of the data source. Additional steps may be included.

In the step where the second dynamic workload processing node dynamically processes the workload, the workload may be dynamically processed based on local information of the data source and information of the second adjacent dynamic workload processing node.

The dynamic workload processing method according to an embodiment of the present invention includes a second dynamic workload processing node where the second dynamic workload processing node includes a second proximate dynamic workload processing node located around the data source based on local information of the data source. A step of obtaining cluster information may be further included.

In the step where the second dynamic workload processing node dynamically processes the workload, the workload may be dynamically processed based on local information of the data source and information of the second cluster.

When there are a plurality of second adjacent dynamic workload processing nodes, in the step of the second dynamic workload processing node dynamically processing the workload, the second dynamic workload processing node is configured to Workloads can be dynamically scheduled so that dynamic workload processing nodes distribute tasks corresponding to task information.

The dynamic workload processing method according to an embodiment of the present invention is based on at least one of whether the first dynamic workload processing node fails to operate and updated local information when there are a plurality of second dynamic workload processing nodes. Based on this, the method may further include replacing one of the plurality of second dynamic workload processing nodes with the first dynamic workload processing node to correspond to the control layer.

According to an embodiment of the present invention, the process of directly transmitting large amounts of data to a cloud server can be omitted, thereby reducing costs due to waiting time that occurs when transmitting to a cloud in the prior art.

According to an embodiment of the present invention, a service can be supported without cost loss in a scenario with a short waiting time.

According to an embodiment of the present invention, by operating based on a Kubernetes container, it is possible to comprehensively solve problems in terms of data integration and processing logic in a way that is not limited by the type of technology on which the edge device operates.

According to an embodiment of the present invention, it is possible to process a big data workload that makes full use of edge server hardware resources.

According to an embodiment of the present invention, efficient big data operations can be processed considering the region/location of the data source.

1 is a conceptual diagram illustrating a service scenario to which a dynamic workload processing method according to an embodiment of the present invention is applied.
Figure 2 is a conceptual diagram illustrating a dynamic workload processing device and system according to an embodiment of the present invention.
Figure 3 is a conceptual diagram illustrating a processing method executed in a dynamic workload processing device and system according to an embodiment of the present invention.
4 is a conceptual diagram illustrating a control layer that is part of a dynamic workload processing device and system according to an embodiment of the present invention.
Figure 5 is a conceptual diagram illustrating an operation layer that is part of a dynamic workload processing device and system according to an embodiment of the present invention.
Figure 6 is a conceptual diagram illustrating a data layer that is part of a dynamic workload processing device and system according to an embodiment of the present invention.
Figure 7 is a conceptual diagram illustrating an inter-layer communication protocol in a dynamic workload processing device and system according to an embodiment of the present invention.
FIG. 8 is an operational flowchart illustrating a dynamic workload processing method executed in a dynamic workload processing device and system according to an embodiment of the present invention.
FIG. 9 is a conceptual diagram illustrating a scenario of dynamic workload scheduling executed in a dynamic workload processing device and system according to an embodiment of the present invention.
FIG. 10 is a conceptual diagram illustrating a scenario of dynamic workload scheduling executed in a dynamic workload processing device and system according to an embodiment of the present invention.
FIG. 11 is a conceptual diagram illustrating an example of a generalized dynamic workload device, dynamic workload system, or computing system capable of performing at least a portion of the processes of FIGS. 1 to 10.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

Terms such as first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, a first component may be named a second component without departing from the scope of the present invention, and similarly, the second component may also be named a first component. The term 'and/or' includes any of a plurality of related stated items or a combination of a plurality of related stated items.

In embodiments of the present application, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.” Additionally, in embodiments of the present application, “one or more of A and B” may mean “one or more of A or B” or “one or more of combinations of one or more of A and B.”

When a component is said to be "connected" or "connected" to another component, it is understood that it may be directly connected to or connected to the other component, but that other components may exist in between. It should be. On the other hand, when it is mentioned that a component is “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

The terms used in this application are only used to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted in an ideal or excessively formal sense unless explicitly defined in the present application. No.

Meanwhile, even if the technology was known before the filing date of this application, it may be included as part of the structure of the invention of this application when necessary, and this will be described herein to the extent that it does not obscure the purpose of the present invention. However, in explaining the configuration of the invention of this application, a detailed description of matters that can be clearly understood by a person skilled in the art as known technology before the date of filing this application may obscure the purpose of the present invention, so excessively detailed description of the known technology is not necessary. Omit it.

For example, technologies for reconfiguring a microservice structure, technologies for implementing a container-based service structure in a Kubernetes environment, etc. may use technologies known before the application of the present invention, and at least some of these known technologies are related to the present invention. It can be applied as an element technology necessary for implementation.

However, the purpose of the present invention is not to claim rights to these known technologies, and the content of the known technologies may be included as part of the present invention within the scope without departing from the spirit of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the attached drawings. In order to facilitate overall understanding when describing the present invention, the same reference numerals are used for the same components in the drawings, and duplicate descriptions of the same components are omitted.

1 is a conceptual diagram illustrating a service scenario to which a dynamic workload processing method according to an embodiment of the present invention is applied.

Referring to FIG. 1, edge servers 210 are placed between the server 240 and the edge device 250, and each of the edge servers 210 is a node between the server 240 and the edge device 250. Computation for data services can be divided.

Users of data services can receive necessary data services through the edge device 250, and can receive data services through a separate user terminal.

Edge device 250 may be a data source. For example, in a data service, operations and processing may be performed on data held by edge device A, and the service may be provided to edge device B on the user's side.

At this time, operations and processing of data held by edge device A may be divided and processed by edge device A, edge device B, and the edge servers 210.

In other words, the edge device 250 can be both a data source and a performing entity that can share part of the performance of the workload.

At this time, in one embodiment of the present invention, a plurality of nodes among the edge servers 210 form one cluster 100, and the nodes within the cluster 100 share roles to process data services, and process data services necessary for data service processing. Workloads can be processed dynamically, and scheduling for workload processing can be performed dynamically.

With the advent of the Internet of Things (IoT), a significant amount of data is being generated in edge networks, and when processing the vast amount of data generated here and providing data services to users, the delay and cost of data transmission through the network are increased. This is pointed out as a representative problem. Transmitting such massive amounts of data to the cloud is a high-cost operation that can cause latency, making low-latency service scenarios unfeasible. To solve these problems, edge computing platform technology has emerged.

Conventional edge computing technology aims to complement cloud computing by utilizing the hardware resources of edge devices. Typically, edge computing devices are placed closer to the edge device 250 of the network to meet latency requirements to reduce processing load.

However, solutions that process data services through edge computing have traditionally been designed to concentrate the workload on the cloud server 240, so the edge devices 250 store the workload on the cloud server in order to process the data services required for each other. The process of scheduling, allocating, and processing workloads had inefficiencies that did not fully utilize the purpose of edge computing.

In the prior art, cloud service providers are traditionally used as a standard solution for processing big data. Cloud services are inherently overly centralized in a few locations (e.g., data centers) around the world.

With the advent of the Internet of Things (IoT), a lot of data is being generated in edge devices 250, and there is a growing need to reduce data transmission delays and costs while performing big data processing in geographically distributed systems.

In general, the closer the edge computing environment is to the edge device 250, the more limited the available resources are, and the smaller the amount of data processed at a given time on a single host, so data operations and streaming are continuously processed and work is required to achieve effective throughput. Load processing is required to occur as quickly as possible.

In order for the workload to be processed as quickly as possible, a workload processing mechanism that takes into account the region/location of the data source, that is, the local information of the data, is supported and required to be processed dynamically.

In comparison, the prior art had the problem of not being able to efficiently process the workload for big data processing due to the increase in demand for data services due to the exponential increase in the number of edge terminals with the advent of the Internet of Things.

In the prior art, the amount of data transmitted to the cloud server 240 for big data processing increased, and problems such as workload overload and long waiting time, as well as increased costs, emerged as problems.

The prior art had a short waiting time, required few resources, and had the problem of incurring unnecessary costs by creating and deleting containers each time, even when processing workloads for frequently occurring services.

In the prior art, there was a problem in that it was difficult to integrate data processing methods developed in different programming languages or technologies.

In order to solve these problems of the prior art, in an embodiment of the present invention, the edge device 250 is configured as a container-based cluster 100, and based on this, systems and devices that can efficiently process workloads can be provided. there is.

In an embodiment of the present invention, the concept of a cluster consisting of edge servers 210 close to the edge device 250 is introduced to increase efficiency in dynamically processing workloads using the cluster.

In an embodiment of the present invention, the edge servers 210 share and perform calculations required for dynamic workload processing and data services, thereby reducing the amount of data and information transmitted to the server 240 and reducing traffic within the entire network. , This makes it possible to provide a dynamic workload processing system, device, and method that can provide data services at low cost.

In order to solve the above-mentioned problems, the embodiment of the present invention considers the region/location information of the data source and uses a long-lived container (i.e., the life cycle is not destroyed at the end of the service and meets the requirements specified by the user). You can perform workload processing using containers (that live and run until you are satisfied) and handle interactions with different data sources through containers.

Embodiments of the present invention may propose a container-based dynamic workload processing system, device, and method that can efficiently process data by considering regional/location information of the data source.

An embodiment of the present invention proposes a dynamic workload processing system, device, and method based on a cluster 100 of edge server 210 nodes that can efficiently process data by considering the region/location information of the data source. You can.

In an embodiment of the present invention, edge server 210 nodes can dynamically reconfigure scheduling. At this time, scheduling may mean reorganizing the execution order and performer of the workload.

In the prior art, since this scheduling process is performed in the main server 240 of the cloud, the necessary information is transmitted to the main server 240, and the execution order and performer of the workload are determined by the main server 240, so the main server (240) 240), the computational load was concentrated and there was a time delay in scheduling decisions.

In an embodiment of the present invention, by setting up a local cluster and managing the edge servers 210 within it, scheduling can be dynamically reconfigured through cooperation among the edge servers 210 without going through the main server 240 of the cloud. That is, the execution order of the workload and/or the execution subject within the cluster of the edge server 210 can be reconfigured.

At this time, the entity performing the workload may be the edge device 250 and/or the edge server 210. The edge device 250 is a data source and an entity that can share part of the workload, and in the service scenario of FIG. 1, it may be a user who receives a data service.

In an embodiment of the present invention, dynamically reconfiguring scheduling may be performed primarily on the edge server 210.

Meanwhile, in an embodiment of the present invention, a cluster can be managed based on long-live containers and dynamic scheduling can be performed. By using long-live containers, the efficiency of dynamic scheduling can be increased.

Figure 2 is a conceptual diagram illustrating a dynamic workload processing device and system according to an embodiment of the present invention.

Referring to FIG. 2, components and interactions within a cluster 100 composed of a plurality of nodes are shown.

The master node 220 may be set in response to the control layer 110, which will be described later. Master node 220 may be one of edge servers 210.

A worker node 230 may be set to correspond to the computation layer 120 and/or the data layer 130, which will be described later. Worker node 220 may be one of edge servers 210.

Which nodes are included in one cluster 100, which nodes are set as the master node 220, and which nodes are set as the worker nodes 230 are determined by the cooperative operation of the nodes within the cluster 100. Can be determined dynamically.

At least some of the properties of the master node 220 and the properties of the worker node 230 may refer to those disclosed in the prior art Kubernetes (K8S) platform. However, it will be readily understood by those skilled in the art that the spirit of the present invention is not limited to embodiments using the Kubernetes platform.

In an embodiment of the present invention, attributes necessary for dynamic setting of the master node 220 and the worker node 230 may be additionally set. The designation of the master node 220 and worker node 230 is not permanent and may change dynamically in response to the situation and needs of the system. Additionally, whether or not each node is included in the cluster 100 and which cluster 100 it is included in are not permanently determined and may change dynamically in response to situations and system needs.

A dynamic workload processing device that dynamically processes a workload for providing a service according to an embodiment of the present invention includes a memory that stores at least one command; and an edge server 210 including a processor that executes at least one command.

At this time, the processor of the edge server 210 obtains task information necessary to support a user service request by performing at least one command, and obtains local information of a data source for processing a task corresponding to the task information, Based on local information, the workload for processing tasks corresponding to task information is dynamically processed.

For convenience of explanation, the cluster to which the target node belongs can be displayed as a first cluster, and clusters other than the first cluster can be displayed as second clusters.

When dynamically processing a workload, the processor may dynamically reconfigure the order of the workload and the scheduling of at least one of the workload performers based on local information. The entity performing the workload may include at least one of each node itself, other nodes in the first cluster, nodes in the second cluster, and the edge device 250.

In obtaining the task information necessary to support the user service request, the processor receives a task information request requesting the task information necessary to support the user service request from the control layer 110 in the first cluster to which the specific node belongs. It is possible to obtain task information necessary to support the user service request by analyzing the user service request corresponding to the task information request.

In obtaining local information of a data source, the processor may obtain local information of a data source for processing a task corresponding to task information from the data layer 130 in the first cluster. The data source may be one of the edge devices 250. A data source may refer to an edge device 250 that holds data necessary to perform a data service.

The processor may obtain information on a first proximate dynamic workload processing node located around the data source based on local information of the data source. The local information of the data source may include the geographical location of the data source and information on nodes located close to the data source. For convenience of explanation, a node selected to share at least a portion of the workload required to perform a data service among nodes located close to the data source may be referred to as a first adjacent dynamic workload processing node. The process of selecting a first proximate dynamic workload processing node among nodes located close to the data source may be performed through cooperative operations of nodes in the first cluster.

The first proximate dynamic workload processing node may or may not be included within the first cluster.

In dynamically processing the workload, the processor may dynamically process the workload based on local information of the data source and information of the first adjacent dynamic workload processing node.

The processor may obtain information of a second cluster including a second proximate dynamic workload processing node located around the data source based on local information of the data source. Among the nodes located close to the data source, a node that shares the processing of the workload and belongs to the second cluster rather than the first cluster may be referred to as a nearby second dynamic workload for convenience of explanation.

When dynamically processing a workload, the processor may dynamically process the workload based on local information of the data source and information of the second cluster.

When there are a plurality of second proximate dynamic workload processing nodes, the processor may dynamically schedule the workload so that the plurality of second proximate dynamic workload processing nodes in the second cluster distribute and process tasks corresponding to the task information. there is. That is, dynamic processing of the workload may be shared by any one of the nodes around the data source, or by a cluster (second cluster) including nodes around the data source. When a plurality of nodes in the second cluster share and process the workload, task processing may be distributed and allocated to a plurality of nodes in the second cluster.

A dynamic workload processing system according to another embodiment of the present invention includes a first dynamic workload processing node corresponding to a control layer 110 in a first cluster; and a second dynamic workload processing node corresponding to the computational layer 120 in the first cluster.

At this time, the first dynamic workload processing node receives the user service request, and the second dynamic workload processing node transmits a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node. And, the second dynamic workload processing node analyzes the user service request corresponding to the task information request to obtain task information necessary to support the user service request, and the second dynamic workload processing node performs the task corresponding to the task information. Local information of the data source for processing is acquired, and the second dynamic workload processing node dynamically processes the workload for processing the task corresponding to the task information based on the local information.

When the second dynamic workload processing node dynamically processes the workload, the order of the workload and the scheduling of at least one of the workload performers may be dynamically reconfigured based on local information.

When the second dynamic workload processing node obtains local information of the data source, data for processing a task corresponding to the task information from the first proximate dynamic workload processing node corresponding to the data layer 130 in the first cluster Local information of the source can be obtained.

The second dynamic workload processing node may obtain information about a second adjacent dynamic workload processing node located around the data source based on local information of the data source.

When dynamically processing the workload, the second dynamic workload processing node may dynamically process the workload based on local information of the data source and information of the second adjacent dynamic workload processing node.

The second dynamic workload processing node may obtain information of a second cluster including a second adjacent dynamic workload processing node located around the data source based on local information of the data source.

The second dynamic workload processing node may dynamically process the workload based on local information of the data source and information of the second cluster.

The second dynamic workload processing node corresponding to the operation layer 120 is, when there are a plurality of second proximate dynamic workload processing nodes, a plurality of second proximate dynamic workload processing nodes in the second cluster correspond to task information. Workloads can be dynamically scheduled to distribute tasks.

When there are a plurality of second dynamic workload processing nodes, one of the plurality of second dynamic workload processing nodes is 1 It may correspond to the control layer 110 by replacing the dynamic workload processing node.

Figure 3 is a conceptual diagram illustrating a processing method executed in a dynamic workload processing device and system according to an embodiment of the present invention.

According to an embodiment of the present invention below, a container-based dynamic workload processing system is disclosed. According to the microservice-based container architecture, user-contributed code and framework-provided components are combined to define and execute user-specifiable workflows. In other words, even if each device or sub-code is implemented in a different programming language or technology, it can be efficiently integrated and managed by using the system according to the embodiment of the present invention because it is executed and combined independently from each other.

Referring to FIG. 3, the cluster 100 of the dynamic workload processing system according to an embodiment of the present invention includes a control layer 110, an operation layer 120, and a data layer 130.

The node of the control layer 110 may transmit a task performance signal to the node of the operation layer 120 (S310).

The node of the computation layer 120 may push data to the node of the data layer 130 (S320).

The nodes of the computation layer 120 may pool data from the nodes of the data layer 130 (S330).

The node of the computation layer 120 may notify the node of the control layer 110 whether the task has been completed (S340).

Data services processed within the cluster 100 may be provided to the cloud server 240 or external users via the external traffic network 340. At this time, the load balancer 350 may partially schedule traffic passing through the external traffic network 340.

FIG. 4 is a conceptual diagram illustrating a control layer 110 that is part of a dynamic workload processing device and system according to an embodiment of the present invention.

The control layer 110 may make decisions related to execution of workflows related to data definition. Workflow refers to a series of sequences for task processing. The control layer 110 includes an orchestrator 112 module, and the orchestrator 112 module may include a strategy container 114 and an API server 116.

Orchestrator 112 may coordinate tasks performed in the computation layer 120. When control layer 110 receives notification from a user or client that new data is available for processing, orchestrator 112 can determine what type of task should be invoked to process the current new data.

The orchestrator 112 determines what type of task to call, and determines the task type and order through the task logic container 126 in the smart task processor 122 module of the operation layer 120, which will be described later. there is. For example, in a workflow consisting of three sequential steps, the above embodiment may include transferring the result of data processing output from the second step to an instance of the third step.

Orchestrator 112 is centralized and a single orchestrator instance can be deployed within a Kubernetes cluster. The corresponding orchestrator instance may perform the role of managing all nodes within the cluster 100. The orchestrator 112 may pursue a single-replica strategy through the strategy container 114. In one embodiment of the present invention, the orchestrator 112 is a stateful service that stores the state in memory, and when a multi-replica strategy is used, synchronization problems with external solutions are eliminated for storing and sharing the state between replicas. Since this occurs, a single replica strategy can be applied. Orchestrator 112 may be deployed through a Kubernetes service.

There may be task processing instances of the same type throughout the system. The strategy container 114 of the orchestrator 112 selects one of the instances to perform data processing and uses a custom routing decision algorithm instead of relying on traditional load balancing algorithms (round robin, random, least connections, etc.). You can. You can then decide on a routing strategy by considering data location, current workload, availability of various resources, and cost. Because synchronous processing of information about all workload targets can result in significant performance degradation and cost issues, processes such as routing decisions can be pre-calculated or asynchronously calculated.

The API server 116 may partially adopt the API server functions and implementation of Kubernetes in the prior art. The API server 116 may receive all requests from distributed hosted nodes and components and provide an HTTP API to enable communication between nodes within the cluster 110 and with external components.

FIG. 5 is a conceptual diagram illustrating an operation layer 120 that is part of a dynamic workload processing device and system according to an embodiment of the present invention.

Computational layer 120 may include a plurality of smart task processor 122 modules. Each of the plurality of smart task processor 122 modules receives order and definition information necessary for task performance from the orchestrator 112 of the control layer 110, configures tasks, and performs workflow. The execution model order in the computation layer 120 can be simplified and implemented as follows.

The smart task processor 122 module may include an agent container 124 and a task logic container 126. The agent container 124 can perform operations, and the task logic container 126 can logically analyze commands provided by the orchestrator 112. These two can always be formed in a sidecar deployment pattern and can be separated into different containers.

Data units may be input to the smart task processor 122.

Data can be logically processed by the task logic container 126. Logical processing may mean the task of analyzing information about the type and/or amount of data. The task logic container 126 can specify the data as input to a specific task defined by the orchestrator 112 and perform the task.

There may be one or multiple results (outputs) for the processed task. Notification of the processing results of the task may be delivered to the orchestrator 126 through the agent container 124.

Agent container 124 can coordinate execution in the context of a single task, such as retrieving input data, invoking processing logic, and processing output data. The agent container 124 may provide the following interface.

The agent container 124 may receive a request (task performance signal) from the orchestrator 112 to process a data unit.

The agent container 124 may retrieve input data by communicating with the data control manager 132 of the data layer 130, which will be described later, according to instructions received from the orchestrator 112.

The agent container 124 may call and perform the task logic container 126 to perform operations based on input data information.

When the operation is completed in the computation layer 120, the output data is delivered (pushed) to the data control manager 132 of the data layer 130 (S320), and the orchestrator 112 is able to process new data. may be notified.

The operation layer 120 included in the embodiment of the present invention can support the function of performing and processing any logic container regardless of programming language or OS type. In the embodiment of the present invention, separate definitions of input data and output data are not required.

FIG. 6 is a conceptual diagram illustrating a data layer 130 that is part of a dynamic workload processing device and system according to an embodiment of the present invention.

The data layer 130 may include a data control manager 132 module and an edge storage 138 module. The data layer 130 may include all components related to data processing. This may include the ability to move data between hosts (nodes), for example to store and retrieve data, and to compute information about required workloads. The data transfer management container 134 is implemented using RDMA (Remote Direct Memory Access), which allows data to be shared between different nodes. RDMA is a method that allows data stored in the memory of one node to be transmitted directly without the intervention of the CPU and OS kernel. Using this method, data can be passed between each step required to execute a workflow. All interactions with edge storage 138 may occur through data control manager 132.

The data transmission management container 134 may be responsible for retrieving input data or output data required to perform a task and delivering it to another layer.

The edge storage 138 is a storage that stores information related to all tasks within the cluster 100 and can be operated in a file system manner. Input data, output data, node information, local information, etc. between tasks can be stored.

The data transfer management container 136 can manage data pushing and pulling. Data retrieved through edge storage 138 is stored in a file, and only data in the file can be uploaded to edge storage 138. The data control manager 132 can be deployed using the DaemonSet concept of Kubernetes. All nodes in the cluster 100 can utilize local storage to store data being processed in a workflow. The data transfer management container 136 may be involved in the process of pushing/pulling data. The data local information management container 136 can capture and store local information about data.

The data local information management container 134 according to an embodiment of the present invention can use a hard link method to more effectively utilize the advantages obtained by using the region/location information of the data source. Since hard linking is generally a faster operation than copying, the time required to move data between directories can be further reduced. Volumes are based on directories in the underlying Node filesystem, and resolution of hard links can be delegated to the Node filesystem, even if different directories are mounted as different volumes in the pod. This allows hard links to cross volumes mounted on different pods. In contrast, symbolic links attempt to resolve a specific path, so they cannot cross different volumes unless all pods mount the same volume under the same path.

The data local information management container 136 can provide physical local information necessary for workload scheduling based on data information of different components. Local information may include the geographical location of the edge device 250 where data is generated, hardware information of the edge device 250, location information between the edge device 250 and other adjacent edge devices, connection information, etc. The data local information management container 136 may have the following characteristics.

The data local information management container 136 may provide local information to both the operation layer 120 and the data layer 130 to provide routing information.

Information about the physical distance to the data source may be calculated in the data local information management container 136, and the calculated information may be shared with all nodes in the cluster 100.

The data local information management container 136 can manage local information based on information fields that are granular enough to collect information on host nodes.

Figure 7 is a conceptual diagram illustrating an inter-layer communication protocol in a dynamic workload processing device and system according to an embodiment of the present invention.

Referring to FIG. 7, the user (client) 260 may transmit a service request to the control layer 110 (S410).

A task information request may be transmitted from the control layer 110 to the operation layer 120 (S420).

Task information that can correspond to the service request is calculated in the computation layer 120, and the task information can be returned from the computation layer 120 to the control layer 110 (S430).

A data request related to task information may be transmitted from the control layer 110 to the data layer 130 (S440). The data layer 130 may return data necessary for task performance to the computation layer 120 (S450). In the computation layer 120, computation for data service may be performed based on data required for task performance. After the operation of the operation layer 120, service information may be transferred from the operation layer 120 to the control layer 110 (S460).

The service support result may be transmitted from the control layer 110 to the user (client) 260 (S470).

The control layer 110 is connected to the computation layer 120 and the data layer 130 and establishes a workflow so that the data needed by the computation layer 120 can be transferred from the data layer 130 to the computation layer 120. You can control it.

The computation layer 120 can calculate task information necessary to perform a service. For example, the operation layer 120 may analyze and recognize that the task information required to perform the data service requested by the user 260 consists of A, B, and C.

The computation layer 120 may return task information A, B, and C required to perform the service to the control layer 110 (S430). The control layer 110 may request data source information from the node of the data layer 130 to check the data source information related to task information A, B, and C, and provide task information A, B, and C. Data necessary for execution can be requested from the data layer 130 (S440).

The data layer 130 may return the data necessary to perform tasks A, B, and C to the operation layer 120 (S450). At this time, the data layer 130 may return local information about the data source needed to perform tasks A, B, and C to the operation layer 120.

The computation layer 120 may dynamically schedule and dynamically process the workload required to perform tasks A, B, and C based on the received data and local information of the data source.

For example, during the initial analysis, it was analyzed that it was optimal to perform tasks in the order of A, B, and C, but after receiving the data and local information of the data source, it was dynamically determined that it was optimal to perform tasks in the order of A, C, and B. You can schedule it.

In Figure 7, an embodiment in which the calculation layer 120 and the data layer 130 are completely separated is shown, but in another embodiment of the present invention, the same node is allocated to the calculation layer 120 and the data layer 130 to perform the functions of the layers. You can also perform .

In an embodiment of the present invention, the nodes of the control layer 110 and the nodes of the operation layer 120/data layer 130 are distinguished, and the nodes of the control layer 110 may not directly perform data processing or operations. .

In one embodiment of the present invention, if the data layer 130 and the operation layer 120 are separated so that the data processing process is performed in the operation layer 120 rather than the data layer 130, the following advantages can be obtained.

According to an embodiment of the present invention, a dynamic workload device and system capable of providing data services can be implemented based on a modular architecture such as microservices. The modular architecture of these microservices can promote reusability so that repetitive tasks can be processed. For example, a container that performs image processing can be reused to process images from multiple data sources, and the same data output can be reused in other computing steps.

Additionally, according to an embodiment of the present invention, there is no problem in the components communicating with each other even if the programming language implementing each data source or edge device 250 is different.

FIG. 8 is an operational flowchart illustrating a dynamic workload processing method executed in a dynamic workload processing device and system according to an embodiment of the present invention.

The dynamic workload processing method according to an embodiment of the present invention is to process a user service request received via a first dynamic workload processing node corresponding to the control layer 110 in the first cluster, to the operation layer in the first cluster. A second dynamic workload processing node corresponding to (120) analyzes (S510) and obtains task information necessary to support a user service request (S520); A second dynamic workload processing node acquiring local information of a data source for processing a task corresponding to the task information (S530); and a step in which the second dynamic workload processing node dynamically processes the workload for processing the task corresponding to the task information based on local information (S540). At this time, the first dynamic workload processing node may be the master node 220 of FIG. 2, and the second dynamic workload processing node may be the worker node 230 of FIG. 2.

In the step where the second dynamic workload processing node dynamically processes the workload, the order of the workload and the scheduling of at least one of the workload performers may be dynamically reconfigured based on local information.

A dynamic workload processing method according to an embodiment of the present invention includes the steps of a first dynamic workload processing node receiving a user service request (S410); It may further include a step (S420) in which the first dynamic workload processing node transmits a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node.

In the step (S530) of the second dynamic workload processing node acquiring local information of the data source, a task corresponding to the task information is selected from the first adjacent dynamic workload processing node corresponding to the data layer 130 in the first cluster. Local information of the data source for processing may be obtained.

The dynamic workload processing method according to an embodiment of the present invention includes the second dynamic workload processing node obtaining information about a second adjacent dynamic workload processing node located around the data source based on local information of the data source. Additional steps may be included.

In the step (S540) in which the second dynamic workload processing node dynamically processes the workload, the workload may be dynamically processed based on local information of the data source and information of the second adjacent dynamic workload processing node.

The dynamic workload processing method according to an embodiment of the present invention includes a second dynamic workload processing node where the second dynamic workload processing node includes a second proximate dynamic workload processing node located around the data source based on local information of the data source. A step of obtaining cluster information may be further included.

In the step (S540) in which the second dynamic workload processing node dynamically processes the workload, the workload may be dynamically processed based on local information of the data source and information of the second cluster.

When there are a plurality of second adjacent dynamic workload processing nodes, in the step (S540) of the second dynamic workload processing node dynamically processing the workload, the second dynamic workload processing node is connected to a plurality of adjacent dynamic workload processing nodes in the second cluster. The workload may be dynamically scheduled so that second adjacent dynamic workload processing nodes distribute and process tasks corresponding to task information.

The dynamic workload processing method according to an embodiment of the present invention is based on at least one of whether the first dynamic workload processing node fails to operate and updated local information when there are a plurality of second dynamic workload processing nodes. Based on this, a step of replacing one of the plurality of second dynamic workload processing nodes with the first dynamic workload processing node and corresponding to the control layer 110 may be further included.

FIG. 9 is a conceptual diagram illustrating a scenario of dynamic workload scheduling executed in a dynamic workload processing device and system according to an embodiment of the present invention.

Referring to FIG. 9, the smart task processor 122 module in the computation layer 120 may be implemented so that multiple instances of the same computation type are executed in parallel.

In Figure 9, an instance for image processing is disclosed as a plurality of instances of the same operation type.

FIG. 10 is a conceptual diagram illustrating a scenario of dynamic workload scheduling executed in a dynamic workload processing device and system according to an embodiment of the present invention.

Referring to FIG. 10, in the computation layer 120, the smart task processor 122 module may have multiple instances of the same computation type distributed across multiple hosts (nodes).

Tasks 1, 2, and 3 to perform data services are connected to hosts (nodes) 1, 2, and ??. It can be allocated to be distributed and processed among n.

FIG. 11 is a conceptual diagram illustrating an example of a generalized dynamic workload processing device, dynamic workload processing system, or computing system capable of performing at least a portion of the processes of FIGS. 1 to 10.

At least some processes of the dynamic workload processing method according to an embodiment of the present invention may be executed by the computing system 1000 of FIG. 11.

Referring to FIG. 11, the computing system 1000 according to an embodiment of the present invention includes a processor 1100, a memory 1200, a communication interface 1300, a storage device 1400, an input interface 1500, and an output. It may be configured to include an interface 1600 and a bus 1700.

The computing system 1000 according to an embodiment of the present invention includes at least one processor 1100 and instructions instructing the at least one processor 1100 to perform at least one step. It may include a memory 1200 for storing. At least some steps of the method according to an embodiment of the present invention may be performed by the at least one processor 1100 loading instructions from the memory 1200 and executing them.

The processor 1100 may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed.

Each of the memory 1200 and the storage device 1400 may be comprised of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory 1200 may be comprised of at least one of read only memory (ROM) and random access memory (RAM).

Additionally, the computing system 1000 may include a communication interface 1300 that performs communication through a wireless network.

Additionally, the computing system 1000 may further include a storage device 1400, an input interface 1500, an output interface 1600, etc.

Additionally, each component included in the computing system 1000 may be connected by a bus 1700 and communicate with each other.

Examples of the computing system 1000 of the present invention include a communication capable desktop computer, laptop computer, laptop, smart phone, tablet PC, and mobile phone. (mobile phone), smart watch, smart glass, e-book reader, PMP (portable multimedia player), portable game console, navigation device, digital camera, DMB (digital) It may be a multimedia broadcasting player, digital audio recorder, digital audio player, digital video recorder, digital video player, PDA (Personal Digital Assistant), etc. .

The operation of the method according to the embodiment of the present invention can be implemented as a computer-readable program or code on a computer-readable recording medium. Computer-readable recording media include all types of recording devices that store information that can be read by a computer system. Additionally, computer-readable recording media can be distributed across networked computer systems so that computer-readable programs or codes can be stored and executed in a distributed manner.

Additionally, computer-readable recording media may include hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Program instructions may include not only machine language code such as that created by a compiler, but also high-level language code that can be executed by a computer using an interpreter, etc.

Although some aspects of the invention have been described in the context of an apparatus, it may also refer to a corresponding method description, where a block or device corresponds to a method step or feature of a method step. Similarly, aspects described in the context of a method may also be represented by corresponding blocks or items or features of a corresponding device. Some or all of the method steps may be performed by (or using) a hardware device, such as a microprocessor, programmable computer, or electronic circuit, for example. In some embodiments, at least one or more of the most important method steps may be performed by such an apparatus.

In embodiments, a programmable logic device (e.g., a field programmable gate array) may be used to perform some or all of the functionality of the methods described herein. In embodiments, a field-programmable gate array may operate in conjunction with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by some hardware device.

Although the present invention has been described with reference to preferred embodiments of the present invention, those skilled in the art may make various modifications and changes to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that it is possible.

Claims (20)

A dynamic workload processing device that dynamically processes workloads to provide services,
A memory that stores at least one command; and
A processor that performs the at least one command;
Including,
The processor performs the at least one instruction,
Obtain necessary task information to support user service requests,
Obtain local information of a data source for processing a task corresponding to the task information,
Dynamically processing a workload for processing a task corresponding to the task information based on the local information,
Dynamic workload processing unit. According to paragraph 1,
The processor,
In dynamically processing the workload,
Dynamically reconfiguring the scheduling of at least one of the order of the workload and the entity performing the workload based on the local information,
Dynamic workload processing unit. According to paragraph 1,
The processor,
In obtaining task information necessary to support the user service request,
Receiving a task information request requesting task information necessary to support the user service request from a control layer in the first cluster,
Analyzing the user service request corresponding to the task information request to obtain task information necessary to support the user service request,
In obtaining the local information of the data source,
Obtaining the local information of the data source for processing a task corresponding to the task information from a data layer in the first cluster,
Dynamic workload processing unit. According to paragraph 1,
The processor,
Obtain information on a first proximate dynamic workload processing node located around the data source based on the local information of the data source,
In dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and information of the first proximate dynamic workload processing node,
Dynamic workload processing unit. According to paragraph 1,
The processor,
Obtaining information on a second cluster including a second proximate dynamic workload processing node located around the data source based on the local information of the data source,
In dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and the information of the second cluster,
Dynamic workload processing unit. According to clause 5,
When the second adjacent dynamic workload processing node is plural,
The processor,
Dynamically scheduling the workload so that the plurality of second adjacent dynamic workload processing nodes in the second cluster distribute and process tasks corresponding to the task information,
Dynamic workload processing unit. a first dynamic workload processing node corresponding to a control layer within the first cluster; and
a second dynamic workload processing node corresponding to a computational layer within the first cluster;
Including,
The first dynamic workload processing node receives a user service request,
The first dynamic workload processing node transmits a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node,
The second dynamic workload processing node analyzes the user service request corresponding to the task information request to obtain task information necessary to support the user service request,
The second dynamic workload processing node acquires local information of a data source for processing a task corresponding to the task information,
The second dynamic workload processing node dynamically processes a workload for processing a task corresponding to the task information based on the local information,
Dynamic workload processing system. In clause 7,
When the second dynamic workload processing node dynamically processes the workload,
Dynamically reconfiguring the scheduling of at least one of the order of the workload and the entity performing the workload based on the local information,
Dynamic workload processing system. In clause 7,
In the second dynamic workload processing node obtaining the local information of the data source,
Obtaining the local information of the data source for processing a task corresponding to the task information from a first proximal dynamic workload processing node corresponding to a data layer in the first cluster,
Dynamic workload processing system. In clause 7,
The second dynamic workload processing node,
Obtain information about a second nearby dynamic workload processing node located around the data source based on the local information of the data source,
In dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and information of the second proximate dynamic workload processing node,
Dynamic workload processing system. In clause 7,
The second dynamic workload processing node,
Obtaining information on a second cluster including a second proximate dynamic workload processing node located around the data source based on the local information of the data source,
In dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and the information of the second cluster,
Dynamic workload processing system. According to clause 11,
When the second adjacent dynamic workload processing node is plural,
The second dynamic workload processing node,
Dynamically scheduling the workload so that the plurality of second adjacent dynamic workload processing nodes in the second cluster distribute and process tasks corresponding to the task information,
Dynamic workload processing system. In clause 7,
When there are multiple second dynamic workload processing nodes,
One of the plurality of second dynamic workload processing nodes replaces the first dynamic workload processing node based on whether the first dynamic workload processing node fails to operate and at least one of the updated local information. Corresponding to the control layer,
Dynamic workload processing system. A second dynamic workload processing node corresponding to a computation layer in the first cluster analyzes a user service request received via a first dynamic workload processing node corresponding to a control layer in the first cluster, and makes the user service request. Obtaining task information necessary to support;
acquiring, by the second dynamic workload processing node, local information of a data source for processing a task corresponding to the task information; and
dynamically processing, by the second dynamic workload processing node, a workload for processing a task corresponding to the task information based on the local information;
Dynamic workload processing method, including. According to clause 14,
In the step of the second dynamic workload processing node dynamically processing the workload,
Dynamically reconfiguring the scheduling of at least one of the order of the workload and the entity performing the workload based on the local information,
How to handle dynamic workloads. According to clause 14,
receiving, by the first dynamic workload processing node, a user service request;
transmitting, by the first dynamic workload processing node, a task information request requesting task information necessary to support the user service request to the second dynamic workload processing node;
It further includes,
In the step of the second dynamic workload processing node acquiring the local information of the data source,
Obtaining the local information of the data source for processing a task corresponding to the task information from a first proximate dynamic workload processing node corresponding to a data layer in the first cluster,
How to handle dynamic workloads. According to clause 14,
obtaining, by the second dynamic workload processing node, information of a second adjacent dynamic workload processing node located around the data source based on the local information of the data source;
It further includes,
In the step of the second dynamic workload processing node dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and information of the second proximate dynamic workload processing node,
How to handle dynamic workloads. According to clause 14,
The second dynamic workload processing node obtaining information of a second cluster including a second adjacent dynamic workload processing node located around the data source based on the local information of the data source;
It further includes,
In the step of the second dynamic workload processing node dynamically processing the workload,
Dynamically processing the workload based on the local information of the data source and the information of the second cluster,
How to handle dynamic workloads. According to clause 18,
When the second adjacent dynamic workload processing node is plural,
In the step of the second dynamic workload processing node dynamically processing the workload,
The second dynamic workload processing node dynamically schedules the workload so that the plurality of second adjacent dynamic workload processing nodes in the second cluster distribute and process tasks corresponding to the task information.
How to handle dynamic workloads. According to clause 14,
When there are multiple second dynamic workload processing nodes,
One of the plurality of second dynamic workload processing nodes replaces the first dynamic workload processing node based on whether the first dynamic workload processing node fails to operate and at least one of the updated local information. Corresponding to the control layer;
Containing more,
How to handle dynamic workloads.