Edge-Native: The New Paradigm For Operating And Developing Edge Apps
summary
Argues cloud-native principles applied to the edge let teams operate thousands of NAT‑hidden sites with a declarative pull model and a central control plane, using Kubernetes-based control plane and site pools. Demonstrates a working system called Great Bear edge management that boots Raspberry Pis, deploys apps (digital signage) and keeps them in sync; Raspberry Pi updates lag but succeed. Describes model compression, model-splitting and Flame federated learning to run/ train AI at the edge and reports example inference speedups from ~1 FPS to ~2.4 FPS.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:07Cloud to edge shift
- 3:13Edge computing landscape
- 4:15Edge device evolution
- 6:40Twelve-factor app principles adapted for edge
- 9:06Scaling edge sites
- 10:10Edge deployment robustness
- 11:37Backing services & ops
- 12:19Observability at edge
- 14:25Kubernetes for edge
- 16:11Live edge deployment demo
- 19:44Event-driven displays
- 23:15AI models for edge
- 26:48ML optimization at edge
- 28:38Federated learning
- 30:23Edge-native benefits
- 31:06Q&A
related talks
12mSMARTER: Experiences with Cloud Native on the Edge
27mBuilding an Edge Computing Platform for Network Services Using Cloud Native Technology [I]
08mKeynote Edge Native is the new Cloud Native Modern Apps at the Edge
26mIntro to KubeEdge: Kubernetes Native Edge Computing Framework
36mEdge Computing in Kubernetes: Best Practices for the Future
11mLevel Setting the Cloud Native Edge: Definitions and Core Principles
34mKubeEdge: Kubernetes Native Edge Computing Framework
32mKubeEdge DeepDive: Extending Kubernetes To the Edge With Real-World Industry Use Case