Research Article
An Efficient Scheduling Strategy for Containers Based on Kubernetes
@INPROCEEDINGS{10.1007/978-3-031-24383-7_18, author={Xurong Zhang and Xiaofeng Wang and Yuan Liu and Zhaohong Deng}, title={An Efficient Scheduling Strategy for Containers Based on Kubernetes}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 18th EAI International Conference, CollaborateCom 2022, Hangzhou, China, October 15-16, 2022, Proceedings, Part I}, proceedings_a={COLLABORATECOM}, year={2023}, month={1}, keywords={Collaborative edge computing Kubernetes Container online scheduling strategy Adaptive weight mechanism Resource utilization}, doi={10.1007/978-3-031-24383-7_18} }- Xurong Zhang
Xiaofeng Wang
Yuan Liu
Zhaohong Deng
Year: 2023
An Efficient Scheduling Strategy for Containers Based on Kubernetes
COLLABORATECOM
Springer
DOI: 10.1007/978-3-031-24383-7_18
Abstract
Container clouds are an important supporting technology for collaborative edge computing, and Kubernetes has become the de facto standard for container orchestration. To solve the problem that the scheduling mechanism of Kubernetes has a single scheduling resource index and is unable to adapt the refined resource scheduling requirements in collaborative edge computing, this paper proposes an efficient multicriteria container online scheduling strategy based on Kubernetes, named E-KCSS. To improve the resource utilization of the cluster, the proposed E-KCSS strategy takes into account the global view of edge nodes and containers. An adaptive weight mechanism based on real-time utilization is proposed to solve the problem that preset Kubernetes weighting coefficients do not meet the individual resource requirements of applications. The experimental results show that compared with the scheduling mechanism of Kubernetes, the deployment efficiency of E-KCSS is improved by 35.22%, the upper limit of container application deployment is increased by 29.82%, and the cluster resource imbalance is reduced by 6.87%, which can make the multi-dimensional resource utilization of the cluster more balanced.
