SMARTER: Experiences with Cloud Native on the Edge
summary
A cloud-native stack built around Kubernetes runs edge workloads as daemon sets, pulls containers from registries, and pushes telemetry via MQTT and Fluent Bit to AWS; communication is edge‑originated with a reverse tunnel for debug and a node-local DNS to resolve local containers. A containerized smarter device manager mediates sensor and actuator access. Prototypes run on Raspberry Pi and NVIDIA Xavier AGX, support VM-backed containers and prototype GPU sharing, and have been stress-tested on thousands of edge nodes.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:36Project Overview
- 2:26Cloud vs Edge
- 4:16Cloud Native
- 5:30Kubernetes
- 7:09Connectivity
- 8:47Smart Device Manager
- 9:45Smart Stack Example
- 10:42Smart Stack Functionality
- 11:32Wrap Up
related talks
32mEdge-Native: The New Paradigm For Operating And Developing Edge Apps
35mKubeEdge – Kubernetes Native Edge Computing Framework
27mBuilding an Edge Computing Platform for Network Services Using Cloud Native Technology [I]
36mEdge Computing in Kubernetes: Best Practices for the Future
34mManaging Edge Nodes With “KubeEdge”
39mIoT Application Running on KubeEdge + ARM Platform
31mIntro to KubeEdge: Kubernetes Native Edge Computing Framework
14mBorn in the Cloud growing at the Edge?