Method for supporting service distribution and load balancing on basis of workload unit weight and reference information

Abstract

A workload weight-based service distribution and load balancing support method is provided. A service management method according to an embodiment of the present invention comprises: packaging a service; calculating the workload weight of the service; determining, on the basis of the calculated workload weight, an image repository in which an image is to be stored; generating an image of the packaged service; and storing the generated image in the determined image repository. Therefore, image storage and service distribution can be effectively managed on the basis of workload weight, and thus a delay problem of a large-capacity service such as an artificial intelligence learning service in a microservice platform operated in a cloud environment can be solved.

Description

How to support service distribution and load balancing based on workload unit weight and baseline information

[0001] The present invention relates to service distribution management, and more particularly, to a method for managing image storage and distribution based on the workload weight of a large-capacity service such as an artificial intelligence learning service.

[0002] Microservices are composed of units that operate in isolated virtual environments using Docker-based containers. Before service operation, microservices undergo various processes, including service packaging, image creation / storage, and service deployment.

[0003] Meanwhile, since the workload weight of a service is proportional to its performance, large-scale AI-related services may experience significant delays. This makes it difficult to operate microservice platforms that require real-time performance and places a significant burden on service developers and operators.

[0004] Accordingly, a plan is required to effectively manage image storage and service distribution to respond to service delays.

[0005] The present invention has been devised to solve the above problems, and the purpose of the present invention is to provide a method for effectively managing workload weight-based image storage and service distribution as a solution to the delay problem of large-capacity services such as artificial intelligence learning services in a microservice platform operated in a cloud environment.

[0006] A service management method according to one embodiment of the present invention for achieving the above object includes the steps of: packaging a service; calculating a workload weight of the service; determining an image storage for storing an image based on the calculated workload weight; generating an image of the packaged service; and storing the generated image in the determined image storage.

[0007] The calculation step may be to calculate the workload weight of the service while packaging the service.

[0008] The decision step may be to determine which cluster to create the image from and where to store the image on another cluster.

[0009] The decision step may be to decide which image repository to store the image in, taking into account the idle space in the image repository.

[0010] The decision step may further consider the distance between the cluster where the node generating the image is located and the cluster where the image repository where the image is to be stored is located to determine where the image is to be stored.

[0011] The decision step may be to decide where to store the image by further considering the similarity between the services provided by the cluster where the image repository is located and the services to be generated as images.

[0012] The service management method according to the present invention may further include a step of distributing a service using a stored image.

[0013] The deployment phase may be performed at a point after the storage phase when service deployment is required.

[0014] The service could be an artificial intelligence learning service.

[0015] According to another aspect of the present invention, a service system is provided, comprising: a cluster that packages a service, generates an image of the packaged service, and stores the generated image in an image storage determined by a decision module; and a decision module that calculates a workload weight of the packaged service and determines a storage in which to store the image based on the calculated workload weight.

[0016] According to another aspect of the present invention, a service management method is provided, comprising: a step of calculating a workload weight of a packaged service; and a step of determining an image repository for storing an image of the packaged service based on the calculated workload weight.

[0017] According to another aspect of the present invention, a service management system is provided, comprising: a processor for calculating a workload weight of a service packaged by a cluster and determining an image repository for storing an image of the packaged service based on the calculated workload weight; and a storage unit for providing storage space required by the processor.

[0018] As described above, according to embodiments of the present invention, by effectively managing image storage and service distribution based on workload weight, it is possible to resolve delay issues in large-capacity services such as artificial intelligence learning services on a microservice platform operating in a cloud environment.

[0019] Figure 1 shows the storage load occurrence situation of the container image repository.

[0020] Figure 2 is a service workload weight-based service management method according to one embodiment of the present invention;

[0021] Figure 3 is an example of determining a container image repository.

[0022] Figure 4 is a hardware configuration of a service management system according to another embodiment of the present invention.

[0023] Hereinafter, the present invention will be described in more detail with reference to the drawings.

[0024] An embodiment of the present invention proposes a workload-weight-based service management method for large-capacity services. This technology effectively manages image storage and service distribution based on workload weight, thereby addressing latency issues for large-capacity services, such as artificial intelligence learning services, on microservice platforms operating in cloud environments.

[0025] Operating microservices in a cloud environment requires an image management environment, such as a container image repository. This environment is a key resource for managing and deploying container images created on nodes within a cluster.

[0026] Accordingly, the image management environment operates on a cluster-by-cluster basis, as illustrated in Figure 1. In other words, if multiple clusters are in operation, each cluster must have at least one container image repository. This is because a large distance between the cluster where the service is deployed and the image management environment can result in significant delays in service deployment and operation.

[0027] Meanwhile, as illustrated in Fig. 1, storage and management of container images without considering the status of the container image repository may cause problems such as causing storage load on the container image repository.

[0028] Accordingly, an embodiment of the present invention proposes a service management method based on the workload weight of a service. This technology stores container images and manages clusters for service deployment based on the workload weight of the service.

[0029] FIG. 2 is a diagram illustrating a flowchart of a service management method based on workload weight according to one embodiment of the present invention.

[0030] For workload weight-based service management, as illustrated, a large-capacity service such as an artificial intelligence learning service is first packaged (S110), and then the workload weight of the service is calculated (S120).

[0031] The size of a container image is determined based on the workload weight of the service. Therefore, in an embodiment of the present invention, the size of the container image is predicted and addressed in advance from the service packaging stage.

[0032] Thereafter, based on the calculated workload weight, a container image repository for storing the container image is determined (S130). The container image repository determined in step S130 may be the container image repository of the cluster where the node generating the container image is located, or it may be the container image repository of a cluster where the node is not located.

[0033] The calculation of the service workload weight in step S120 and the determination of the container image repository in step S130 can be performed by the image repository determination module, as illustrated in FIG. 3. The image repository determination module can be provided for each cluster, but can also be located on a separate management server to manage image repository determination for all clusters.

[0034] When determining a container image repository, the image repository decision module considers the size of the container image to be created and the available free space in the container image repository. Furthermore, the image repository decision module considers the distance between the cluster where the node generating the container image is located and the cluster where the container image repository will store the container image.

[0035] Additionally, the image repository decision module can additionally consider the similarity between the services provided by the node generating the container image and the services provided by the cluster where the container image repository where the container image will be stored is located.

[0036] Figure 3 shows a situation where the image repository decision module decides to store container images to be generated on nodes of cluster A in the container image repository of cluster B.

[0037] Afterwards, the node creates a container image of the service packaged in step S110 (S140), and stores the image created in step S140 in the container image repository of the cluster determined in step S130 (S150).

[0038] Afterwards, services can be deployed to the cluster's nodes using container images stored in the container image repository (S160). Step S160 is optional. That is, only steps up to S150 may be performed, and step S160 may not be performed.

[0039] If implemented, service deployment can be performed at any time after container image storage. In this case, if service deployment is performed, pre-stored container images are used, enabling seamless service provision.

[0040] FIG. 4 is a diagram illustrating the hardware configuration of a service management system according to another embodiment of the present invention. The service management system according to the embodiment of the present invention refers to the image storage determination module illustrated in the center of FIG. 3.

[0041] The container scheduling system according to the embodiment of the present invention can be implemented as a cloud / server system including a communication unit (210), a processor (220), and a storage unit (230) as illustrated.

[0042] The communication unit (210) is a communication interface for connection with external networks or devices. The processor (220) manages service distribution by calculating the workload weight of the aforementioned service and determining the container image repository. The storage unit (230) provides the storage space necessary for the processor (220) to function and operate.

[0043] So far, we have described in detail a preferred embodiment of a method for supporting workload weight-based service distribution and load balancing.

[0044] In the above embodiment, by effectively managing image storage and service distribution based on workload weight, the delay problem of large-scale services such as artificial intelligence learning services on a microservice platform operating in a cloud environment can be resolved.

[0045] Meanwhile, it goes without saying that the technical idea of ​​the present invention can also be applied to a computer-readable recording medium containing a computer program that performs the functions of the device and method according to the present embodiment. In addition, the technical idea according to various embodiments of the present invention can be implemented in the form of computer-readable code recorded on a computer-readable recording medium. The computer-readable recording medium can be any data storage device that can be read by a computer and store data. For example, the computer-readable recording medium can be a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical disk, a hard disk drive, etc. In addition, the computer-readable code or program stored on the computer-readable recording medium can be transmitted through a network connected between computers.

[0046] In addition, although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications can be made by a person having ordinary skill in the art to which the present invention pertains without departing from the gist of the present invention as claimed in the claims, and such modifications should not be understood individually from the technical idea or prospect of the present invention.

Claims

1. Step of packaging the service; Step for calculating the workload weight of the service; A step of determining an image repository to store the image based on the calculated workload weight; Step 1: Create an image of the packaged service; A service management method, characterized by comprising a step of storing the generated image in a determined image storage.

2. In claim 1, The calculation steps are: A service management method characterized by calculating a workload weight of a service while packaging the service.

3. In claim 1, The decision stage is, A service management method characterized by being able to determine an image repository of a cluster other than the cluster in which an image is to be created.

4. In claim 3, The decision stage is, A service management method characterized by determining an image storage for storing images by considering idle space in the image storage.

5. In claim 3, The decision stage is, A service management method characterized in that the storage for storing an image is determined by further considering the distance between the cluster where the node for generating the image is located and the cluster where the image storage for storing the image is located.

6. In claim 3, The decision stage is, A service management method characterized in that the storage for storing an image is determined by further considering the similarity between the services provided by the cluster where the service to be generated as an image and the image storage for storing the image are located.

7. In claim 1, A service management method, characterized by further comprising a step of distributing a service with a saved image.

8. In claim 7, The distribution phase is, A service management method characterized in that it is performed at a point in time when service distribution is required after the storage phase.

9. In claim 1, The service is, A service management method characterized by being an artificial intelligence learning service.

10. A cluster that packages services, creates images of the packaged services, and stores the created images in an image repository determined by the decision module; and A service system characterized by including a decision module that calculates a workload weight of a service to be packaged and determines a storage in which to store an image based on the calculated workload weight.

11. Step for calculating the workload weight of the service being packaged; A service management method, characterized by including a step of determining an image repository in which to store an image of a packaged service based on a calculated workload weight.

12. A processor that calculates the workload weight of the service packaged by the cluster and determines the image repository where the image of the packaged service will be stored based on the calculated workload weight; and A service management system, characterized by including a storage unit that provides storage space required by a processor.

Images