The document presents an overview of applying cloud-native architecture to edge computing using OpenYurt, highlighting its benefits such as edge autonomy, interoperability with Kubernetes, and support for service meshes like Istio. It discusses goals for collaborative applications and a dynamic data pipeline, along with feedback received from trials and challenges related to edge optimization and community support. The document emphasizes the relevance of OpenYurt in enhancing edge capabilities and ensuring seamless integration with existing Kubernetes infrastructure.

1. Applying cloud native architecture
to edge with OpenYurt
Tomoya Fujita
Sony Group Corporation, R&D Center, US Laboratory
Jan.10th.2023

2. Agenda
• Introduction
• Background
• Goal
• Why OpenYurt?
• Trial Overview
• Feedback
• Challenges

3. Self Introduction
• Tomoya Fujita
• Sony R&D Center, US Laboratory
• Senior Staff Engineer
• ROS TSC (Technical Steering Committee)
• Github / Linkedin
• OSS Related Talks / Activities
o ROSCon 2022 Organization and Programing Committee
o ROS World 2021 Content Filtered Topic
o KubeCon EU 2021 Edge Day Robotics Edge Cluster System
• Colleagues (Based in Shanghai)
o Barry Xu ( BarryXu-2000@wechat )
o Feng Gao ( 15618992861@wechat )
o Yinzhe Wu ( Windrow14@wechat )

4. Sony RDC Introduction
• Xxx
• xxx

5. Sony RDC Introduction
• Xxx
• xxx

6. Background
• Edge Device Getting Matured
• Distributed / Connected System and Application
• Application Developer Friendly
• Platform Agnostic Application Lifecycle

7. Goals
• Highly Collaborative Application
• Dynamic and Reconfigurable Data Pipeline
• Large-Scale Distributed System and Application
• Transparent platform for Cloud and Edge
• Load-Balancing as entire infrastructure
• Zero-Trust Security

8. Goals
User / Environment
Hardware
Sensor Display Actuator Audio
IN OUT
Sensing
Recognition
Modeling
Intelligent Processing
Rendering
Control
Navigation
Planning
Edge
Cloud
Edge
Cloud
Transparent Orchestration
Usecase-A
Usecase-B

9. Why OpenYurt?
• Provides Data-Plane Tunneling
• Interoperability with Kubernetes mainline
• Edge autonomy
• Can work with Istio Service Mesh
• CNCF Sandbox Project / OSS Activities

10. PoC Overview
Cloud
Edge
Device-A Device-B Device-C
Instance-A Instance-B Instance-C
Kubernetes
Control Plane
Istio
Control Plane
OpenYurt
(Cloud)
OpenYurt
(Edge)
OpenYurt
(Edge)
OpenYurt
(Edge)
Kubernetes
(kubelet)
Application
Container
Application
Container
Application
Container
Application
Container
Application
Container

11. Feedback
• Easy to extend “Cloud Native” to edge computing
• CNI (flannel/cilium/weave) works w/o problems
• Some CNI depends on Kubernetes API
• Interoperability with Kubernetes
• Compatibility with Kubernetes Mainline
• Service Mesh (Istio) also can be supported
• Tunneling Problem (Solved)
• OpenYurt needs to know if tunneling required or not.
• Detail: https://github.com/openyurtio/openyurt/issues/1024

12. Challenges
• Edge Optimization / Reduce Resource Consumption
• Includes network communication cost reduction
• Support up-to-date Kubernetes
• How early we can catch up with Kubernetes?
• More community support required.
• No change required to Kubernetes
• e.g) NodeLifecycle Controller
• Not OpenYurt but Kubernetes
• Device Abstraction / Device Plugin Enhancement

13. Kubernetes Edge IoT WG
Runtime TAG GitHub
https://github.com/cncf/tag-
runtime/blob/master/wg/iot-
edge.md
#wg-iot-edge on K8s
Slack
Edge Native Application
Principles White Paper
Meetings 9AM PT bi-
weekly
https://www.cncf.io/calendar/

14. THANKS