Centralized, distributed or both? - Architectures for Kube...
summary
Edge architectures split into two concrete options: centralized control plane where control plane runs in the core and edges host workloads only (losing the ability to schedule new pods on disconnect), and distributed control plane where full Kubernetes control runs per site for autonomy. Examples mapped to each: k3s and KubeEdge-style setups favor slim centralized deployments, StarlingX and KubeFed support fully distributed/federated clusters, Metal³ (Redfish) handles remote provisioning and Submariner provides cross-cluster networking. Hybrid OpenStack+Kubernetes mixes these tradeoffs.
AI-written summary · about
chapters (6)
- 0:00Intro
- 0:41Edge Computing
- 3:58Architecture Models
- 7:09Projects
- 14:04Other projects
- 17:44Call for action
related talks
29mKubernetes as a Universal Control Plane
34mDistributed applications and Kubernetes
39mBuilding Distributed Systems with Kubernetes
21mChaos Engineering Practice Under Ultra-large-scale Cloud Native Edge Computing
42mThe Challenges and Solutions of Open Edge Infrastructures
37mKubernetesception: How to run a 1000 control planes
43mSystems @Scale Summer Q&A Week 2
31mAn Adventure in Multi Cloud with Kubernetes and Distributed Data