Bridging Baremetal and Virtualized Clouds: Operating Telco-Grade Kubernetes at Scale
summary
Ericsson describes a Kubernetes-native platform that unifies bare metal, OpenStack (via CAPO) and KubeVirt virtualization using Cluster API and Metal3.io, running in production across tens of thousands of servers. Operators use declarative YAMLs, image‑based provisioning and rolling reprovisioning with zero SSH, support failure domains, DHCP‑less boot, OCI images and firmware upgrades; single bare‑metal nodes provision in ~15–100 minutes so a minimal HA cluster appears in about an hour, and upgrades are performed by reprovisioning rather than ad hoc scripting.
AI-written summary · about
chapters (10)
- 0:00Introduction to telco cloud challenges
- 3:16Telco workload requirements across sites
- 7:17Cloud-native telco infrastructure approach
- 11:14Cluster API and Metal³ overview
- 15:50Metal³ bare metal provisioning workflow
- 20:48Metal³ features and production scale
- 25:25Key learnings from operating at scale
- 26:29Q&A on provisioning speed and firmware updates
- 30:02Q&A on network topology and legacy VMs
- 31:47Q&A on OS choices and LLDP detection
related talks
29mManaging Telco Infrastructure and Applications at Scale: An Open Source Approach
19mBridging CNCF and OpenInfra: Kubernetes as the Unified Control Plane for Sovereign Clouds
36mBare metal Kubernetes on mixed x64 and ARM64
20mMetalkube: Deploy Kubernetes on Bare Metal
28m03 Kubernetes Evolution, from the Hard Way to Managed Multi-cloud
40mOperating a Self-Healing Bare-Metal Kubernetes Platform at Global Scale
20mKubernetes at the Edge for Telcos
35mBuilding a Kubernetes on Bare-Metal Cluster