KubeMSEN: Extend KubeEdge to Support Metadata Sharing among Edge Nodes

Abstract

KubeEdge is a popular cloud-native edge computing platforms, enabling collaborative management of cloud-edge computing resources. Metadata is control data of KubeEdge and transmitted massively between cloud and edge when microservice based application is deployed to edge and deletion from edge. However, KubeEdge currently lacks of optimization to deployment and deletion of microservice based application, resulting in high cloud-edge metadata traffic. In this paper, we propose KubeMSEN, extending KubeEdge to support metadata sharing among edge nodes. We firstly use distributed Raft election algorithm to elect leader among edge nodes. Then leader is responsile for sharing metadata among edge nodes to reduce cloud-edge metadata traffic. KubeMSEN is completely cloud-native and relatively simple, only have one additional module based on KubeEdge. KubeMSEN can be also implemented based on other cloud-native platform such as Kubernetes and Openyurt. We tested KubeMSEN in real KubeEdge cluster, and the experimental results show that KubeMSEN can effectively reduce cloud-edge metadata traffic of application deployment and deletion by 33% and 42% respectively compared with native KubeEdge.