OpenStack brings Kubernetes to the Edge
summary
A push-button Kubernetes platform running on OpenStack deploys to many small sites, recommending one cluster per site and avoiding stretched clusters; live migration works inside an AZ but cross-AZ moves are cold. Images and container registries are cached or pre-imported to edge sites; Cinder and Manila with CSI expose zone-aware PVCs for local storage. Network patterns favor spine-leaf and routed provider networks today and plan BGP L3 stretching; expect ~100 ms RTT guidance, installer automation via TripleO/IPI, and operational gotchas like Neutron latency and metadata cache stampedes.
AI-written summary · about
chapters (15)
- 0:00Introduction to speakers
- 1:07Agenda and edge computing overview
- 3:59Geo-distributed Edge architecture
- 7:14Data center edge architecture
- 10:29Compute and storage under the hood
- 13:23Spine-leaf networking benefits
- 18:00Storage optimization for edge
- 24:43Roadmap and future features
- 30:23Lessons learned from field deployments
- 39:39Bloomberg Compute Cloud scaling challenges
- 43:52Network architecture evolution
- 51:36Operationalizing large OpenStack clouds
- 57:56Nova metadata API cache stampedes
- 1:02:36Disaggregating control plane
- 1:06:25Unsolved problems and future work
related talks
45mBring OpenStack & Ceph to the Edge with Control Plane Services Managed by Kubernetes
10mKubernetes: Enabling Edge Workloads
07mKeynote: Pushing the Boundaries of Open Source for a Brand New Edge
24mDemo of Kubernetes at the Edge for AI Model Deployments
20mKubernetes at the Edge for Telcos
26mRun your Kubernetes cluster on OpenStack in Production
31mFrom Region To Multi-AZ: Building Resilient Cloud Infrastructure With OpenStack, Kubernetes, and OVN
31mCloud Provider OpenStack Intro & Update