Enabling Edge Computing Using Container Orchestration and Software Defined Networking

Abstract

With software-defined wide-area networks (SD-WAN) being increasingly adopted, and Kubernetes becoming the de-facto container orchestration tool, the opportunities for deploying edge-computing applications running over a SD-WAN scenario are vast. In this context, a service discovery function will help developing a dynamic infrastructure where clients are able to seek and find particular services. Service discovery also enables a self-healing network capable of detecting the unavailable services. Most of the research in the service discovery field focuses in the discovery of cloud-based services over software-defined networks (SDN). A lack of research in containerized service discovery over SD-WAN is evident. In this thesis, an in-house service discovery solution that works alongside a container orchestrator for allowing an improved traffic handling and better user experience through containerized service discovery and service requests redirection is developed. First, a proof-of-concept SD-WAN topology was implemented alongside a Kubernetes cluster and the in-house service discovery solution. Next, the implementation's performance is tested based on the time required for discovering whether a service has been created, updated or removed. Finally, improvements in node distance computation, local breakout support and the usage of data plane programmability are discussed.