Leveraging Kubernetes in Edge-Native Cable Access Convergence

Abstract

Public clouds provide infrastructure services and deployment frameworks for modern cloud-native applications. As the cloud-native paradigm has matured, containerization, orchestration and Kubernetes have become its fundamental building blocks. For the next step of cloud-native, an interest to extend it to the edge computing is emerging. Primary reasons for this are low-latency use cases and the desire to have uniformity in cloud-edge continuum. Cable access networks as specialized type of edge networks are not exception here. As the cable industry transitions to distributed architectures and plans the next steps to virtualize its on-premise network functions, there are opportunities to achieve synergy advantages from convergence of access technologies and services. Distributed cable networks deploy resource-constrained devices like RPDs and RMDs deep in the edge networks. These devices can be redesigned to support more than one access technology and to provide computing services for other edge tenants with MEC-like architectures. Both of these cases benefit from virtualization. It is here where cable access convergence and cloud-native transition to edge-native intersect. However, adapting cloud-native in the edge presents a challenge, since cloud-native container runtimes and native Kubernetes are not optimal solutions in diverse edge environments. Therefore, this thesis takes as its goal to describe current landscape of lightweight cloud-native runtimes and tools targeting the edge. While edge-native as a concept is taking its first steps, tools like KubeEdge, K3s and Virtual Kubelet can be seen as the most mature reference projects for edge-compatible solution types. Furthermore, as the container runtimes are not yet fully edge-ready, WebAssembly seems like a promising alternative runtime for lightweight, portable and secure Kubernetes compatible workloads.