Implementing a Network-Aware Kubernetes Scheduler on top of a Mesh Network

A recent trend observed in the Kubernetes world is trying to deploy a Kubernetes cluster entirely or partially to the edge. When further looking into these edge environments, it is often noted that wireless technologies are used. Kubernetes is not designed to support these wireless network setups and will have to be extended to run smoothly on these networks. One of the most used network setups in edge environments is a wireless mesh network. Therefore the main focus will be on running Kubernetes on top of a wireless mesh network. Other mesh networks will also be supported with a minimal set of changes needed. The main goal of this paper is to list all the problems encountered when running Kubernetes on top of a mesh network and provide a framework of components and extensions to solve these problems. The majority of these components will be implemented, and demonstrations show that these components are able to solve most of the listed problems when used in a setup with some predefined restrictions. When deploying a demo application with the proposed solution, the required bitrate is observed 97 percent of the time compared to 42 percent with the native Kubernetes scheduler.