Akri: Making IoT Devices Accessible to Your Edge Kubernetes Clusters
summary
Akri exposes physical IoT devices as Kubernetes extended resources and automatically deploys per-device broker pods (example: a frame server using an RTSP URL) so workloads can request devices like any GPU. Discovery handlers (examples: onvif, udev, opcua) run as separate pods behind a gRPC interface so handlers can be written in any language despite Akri itself being built in Rust. Features include automatic eviction when a device disappears, a capacity setting to control sharing, Helm installation, and an end-to-end demo.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:33Kubernetes on the Edge
- 4:19Scenario: Smart Farm
- 5:04Scenario without Akri
- 8:27Scenario Simplified with Akri
- 11:24Akri Architecture
- 12:48Akri Flow
- 15:49Installing with Helm
- 16:13Discovery Handlers as Pods
- 17:30Discovery Handler Interface
- 18:49Akri Functionality Recap
- 20:42Roadmap
- 21:18Together, we can electrify the edge!
related talks
04mAkri: Using non-Kubernetes devices in Kubernetes
27mBridge Your IoT Leaf Devices to Local Clusters with Ease Using Akri...
42mLeaving no Leaf Device Behind: At the Frontier of Edge Computing
09mMaking IoT Simple: Leveraging Akri & Nephio in a Hybrid Cloud En...
24mFrom Sensors to Servers: Efficient Edge Computing with Akri and WebAssembly
39mKubernetes IoT Edge Working Group: Edge Device Onboa...
21mConnecting to devices at the edge with minimal footprint using AKS Edge Essentials and Akri on WASMs
35mCloud Native IoT: OTA Updates and Device Repurposing With K8s