• Journal of Korea Society of Industrial Information Systems
• /
• v.26 no.1
• /
• pp.11-19
• /
• 2021

The emerging Fifth Generation (5G) network technology brings us a new demand for low latency services. However, it may not be possible for long-distanced cloud computing servers to support users with satisfactory low latency services. For this reason, Multi-access Edge Computing (MEC) technology are gaining attraction since it is designed to provide low latency services to users by placing cloud computing resources to base-stations or mobile switching centers nearby users. Accordingly, it is necessary to verify the deployed containers on the MECs are reliable enough to provide low latency services empirically. For the purpose, we develop a testing tool to verify the reliability as well as network resources status of running MECs by deploying containers on the MECs in a Kubernetes environment.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.91-101
• /
• 2024

The utilization of container virtualization technology ensures the consistency and portability of data-intensive and memory volatile workflows. Kubernetes serves as the de facto standard for orchestrating these container applications. Cloud users often overprovision container applications to avoid container restarts caused by resource shortages. However, overprovisioning results in decreased CPU and memory resource utilization. To address this issue, oversubscription of container resources is commonly employed, although excessive oversubscription of memory resources can lead to a cascade of container restarts due to node memory scarcity. Container restarts can reset operations and impose substantial overhead on containers with high memory volatility that include numerous stateful applications. This paper proposes a technique to mitigate container restarts in a memory oversubscription environment based on Kubernetes. The proposed technique involves identifying containers that are likely to request memory allocation on nodes experiencing high memory usage and temporarily pausing these containers. By significantly reducing the CPU usage of containers, an effect similar to a paused state is achieved. The suspension of the identified containers is released once it is determined that the corresponding node's memory usage has been reduced. The average number of container restarts was reduced by an average of 40% and a maximum of 58% when executing a high memory volatile workflow in a Kubernetes environment with the proposed method compared to its absence. Furthermore, the total execution time of a container workflow is decreased by an average of 7% and a maximum of 13% due to the reduced frequency of container restarts.

• The Journal of Korea Robotics Society
• /
• v.17 no.2
• /
• pp.118-123
• /
• 2022

We implemented a real-time cloud robotics application by offloading robot navigation engine over to 5G Mobile Edge Computing (MEC) sever. We also ran a fleet management system (FMS) in the server and controlled the movements of multiple robots at the same time. The mobile robots under the test were connected to the server through 5G SA network. Public 5G network, which is already commercialized, has been temporarily modified to support this validation by the network operator. Robot engines are containerized based on micro-service architecture and have been deployed using Kubernetes - a container orchestration tool. We successfully demonstrated that mobile robots are able to avoid obstacles in real-time when the engines are remotely running in 5G MEC server. Test results are compared with 5G Public Cloud and 4G (LTE) Public Cloud as well.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.07a
• /
• pp.45-48
• /
• 2020

최근 마이크로서비스 아키텍처(Microservice Architecture)를 도입하는 프로젝트가 많아짐에 따라 컨테이너 오케스트레이션 도구인 쿠버네티스의 필요성이 증대되고 있다. 쿠버네티스 환경을 구성하고 서비스를 테스트 및 배포하기 위해서는 클러스터링을 하고 yaml파일을 작성하고 적용하여 자원을 생성하는 등의 작업이 필요하다. 이렇게 반복되는 작업을 자동화하기 위해 기존에는 쉘스크립트를 사용하는데, 이를 사용하기 위해서는 쿠버네티스 및 인프라에 관한 기초지식이 있어야 하는 문제점이 있다. 본 논문에서는 마이크로서비스를 위한 웹 기반의 배포 시스템을 제안한다. 웹을 통해 배포하고 테스트할 수 있어 기초지식이 없어도 사용할 수 있다. 웹에서 프로젝트 개발자로부터 도커 이미지와 사용할 포트번호 및 레플리카 개수를 입력받아 배포를 요청하면 쿠버네티스 환경에 배포하고, 필요한 자원들이 배포가 완료되면 엔드포인트를 제공한다. 그리고 배포된 서버의 엔드포인트에 요청을 전송하고 응답이 오는 것을 확인할 수 있어 활용성이 높다. 이 시스템은 쿠버네티스를 이용하여 복수 개의 파드(Pod)를 유지하여 가용성을 보장하고 빠르게 배포가 가능하게 한다. 또한 이미지가 업데이트되면 웹에서 요청하여 클러스터에 새 버전의 이미지로 배포한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.772-773
• /
• 2023

클라우드 네이티브(Cloud Native) 환경인 쿠버네티스(Kubernetes)에서 컨테이너(Container)는 가볍고 배포 주기가 빠르기 때문에, 쿠버네티스에서 상태를 지속적으로 모니터링하는 시스템이 필요하다. 쿠버네티스는 파드(Pod)를 기본 단위로하여 다수의 컨테이너를 관리·감독한다. 쿠버네티스 모니터링을 위해 프로메테우스는 주기적으로 메트릭을 수집하며 많은 양의 정보를 빠르게 검색한다. 이를 활용하여 쿠버네티스를 모니터링 한다.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.6
• /
• pp.293-301
• /
• 2023

With the advancement of artificial intelligence and its various use cases, accessing it through edge computing environments is gaining traction. However, due to the nature of edge computing environments, efficient management and optimization of clusters distributed in different geographical locations is considered a major challenge. To address these issues, this paper proposes a centralization and automation tool called KubEVC-Agent based on Kubernetes. KubEVC-Agent centralizes the deployment, operation, and management of edge clusters and presents a use case of the data transformation for optimizing intra-cluster communication. This paper describes the components of KubEVC-Agent, its working principle, and experimental results to verify its effectiveness.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.108-111
• /
• 2024

클라우드 사용이 보편화 되고 확대됨에 따라, 서비스를 유연하게 확장 및 축소하여 신속하게 시장의 수요에 대응할 수 있는 PaaS(Platform-as-a-Service) 형태의 서비스가 많은 기업에서 각광받고 있다. 그리고 이러한 PaaS 형 서비스의 핵심이 되는 기술인 컨테이너(Container)와 컨테이너 관리를 효율화 해주는 쿠버네티스(Kubernetes)가 실질적인 표준으로 사용되고 있다. 이때 쿠버네티스 기반의 환경에서 서비스 어플리케이션은 다양한 구성사례가 존재하나, DB 는 아직 안정성 및 데이터 정합성 등을 이유로 베어메탈(Baremetal)이나 VM(Virtual Machine)을 기반으로 구성하고 있는 상황이다. 그러나, 인프라 구성 및 운영에 있어서도 파드(Pod) 형태의 DB 구성은 베어메탈 및 VM 대비 장점이 존재한다고 생각하여 본 실험을 수행하였다. 본 논문에서는 서비스 응답시간 및 자원 사용의 효율성 측면에서 VM 기반의 DB 와 쿠버네티스 파드 기반의 DB 에 각각 트래픽을 발생시켜서 비교한 결과와 시사점을 제시한다.

• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.7
• /
• pp.205-216
• /
• 2022

One of the many tools used for distributed deep learning training is Kubeflow, which runs on Kubernetes, a container orchestration tool. TensorFlow jobs can be managed using the existing operator provided by Kubeflow. However, when considering the distributed deep learning training jobs based on the parameter server architecture, the scheduling policy used by the existing operator does not consider the task affinity of the distributed training job and does not provide the ability to dynamically allocate or release resources. This can lead to long job completion time and low resource utilization rate. Therefore, in this paper we proposes a new operator that efficiently schedules distributed deep learning training jobs to minimize the job completion time and increase resource utilization rate. We implemented the new operator by modifying the existing operator and conducted experiments to evaluate its performance. The experiment results showed that our scheduling policy improved the average job completion time reduction rate of up to 84% and average CPU utilization increase rate of up to 92%.

• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.3
• /
• pp.73-82
• /
• 2022

Auto-scaling is one of the most important functions for cloud computing technology. Even if the number of users or service requests is explosively increased or decreased, system resources and service instances can be appropriately expanded or reduced to provide services suitable for the situation and it can improves stability and cost-effectiveness. However, since the policy is performed based on a single metric data at the time of monitoring a specific system resource, there is a problem that the service is already affected or the service instance that is actually needed cannot be managed in detail. To solve this problem, in this paper, we propose a method to predict system resource and service response time using a multivariate time series analysis model and establish an auto-scaling policy based on this. To verify this, implement it as a custom scheduler in the Kubernetes environment and compare it with the Kubernetes default auto-scaling method through experiments. The proposed method utilizes predictive data based on the impact between system resources and response time to preemptively execute auto-scaling for expected situations, thereby securing system stability and providing as much as necessary within the scope of not degrading service quality. It shows results that allow you to manage instances in detail.

• International journal of advanced smart convergence
• /
• v.10 no.4
• /
• pp.278-288
• /
• 2021

Recently, IoT systems are cloud-based, so that continuous and large amounts of data collected from sensor nodes are processed in the data server through the cloud. However, in the centralized configuration of large-scale cloud computing, computational processing must be performed at a physical location where data collection and processing take place, and the need for edge computers to reduce the network load of the cloud system is gradually expanding. In this paper, a cluster system consisting of 6 inexpensive Raspberry Pi boards was constructed to perform fast data processing. And we propose "Kubernetes cluster system(KCS)" for processing large data collection and analysis by model distribution and data pipeline method. To compare the performance of this study, an ensemble model of deep learning was built, and the accuracy, processing performance, and processing time through the proposed KCS system and model distribution were compared and analyzed. As a result, the ensemble model was excellent in accuracy, but the KCS implemented as a data pipeline proved to be superior in processing speed..