How OpenYurt Enables Scalable Cloud‑Native Edge Computing with Kubernetes

Edge Computing Overview

Edge computing moves compute resources closer to data sources and end‑users to reduce latency and cost, a trend driven by 5G, IoT, video streaming, CDN, and other workloads across industries such as automotive, agriculture, energy, and transportation.

Trends in Edge Computing

AI, IoT and edge computing are converging, leading to increasingly diverse, large‑scale, and complex workloads at the edge.

Edge becomes an extension of cloud computing, requiring decentralized infrastructure, autonomous edge facilities, and cloud‑edge hosting capabilities.

The proliferation of 5G, IoT, and video services fuels rapid growth of edge deployments, exemplified by the pandemic‑driven surge in live streaming and online education.

Challenges for Cloud‑Edge Integration

Traditional Kubernetes workload management cannot meet the scale and complexity of edge workloads.

Public‑network‑based cloud‑edge connections are unstable, threatening edge service reliability.

Edge nodes often sit behind firewalls, resulting in one‑way connectivity that hampers native Kubernetes monitoring and management.

Heterogeneous edge hardware (Linux/Windows, amd64/arm/arm64) makes standardized support difficult.

OpenYurt Architecture

OpenYurt is a non‑intrusive, CNCF‑graduated project that extends native Kubernetes with additional components while preserving upstream compatibility. The architecture consists of a cloud side with standard Kubernetes components (blue boxes) and an edge side with OpenYurt components (orange boxes) connected via public networks.

Key Capabilities of OpenYurt

Unitized Management : Groups edge nodes for batch operations, fine‑grained traffic control, and policy enforcement via the yurt‑app‑manager component.

Edge Autonomy : Ensures workloads continue running when cloud‑edge links are unstable, provided by yurt‑controller‑manager and YurtHub.

Cloud‑Edge Collaboration : Enables full Kubernetes tooling (e.g., kubectl logs, Prometheus) over one‑way networks using tunnel‑server / tunnel‑agent.

Seamless Conversion : Allows one‑click transformation between standard Kubernetes clusters and OpenYurt clusters via the yurtctl tool, verified on Minikube, Kubeadm, and ACK.

Heterogeneous Resource Support : Supports amd64, arm, and arm64 architectures across edge nodes.

Real‑World Use Cases

Case 1 – Hema Fresh : Uses OpenYurt to unify heterogeneous edge compute (including GPU nodes) for a “people‑goods‑store” digital transformation, achieving elastic resource allocation and mixed‑workload flexibility.

Case 2 – Transportation Video Cloud : Deploys cloud‑edge collaboration to bring centralized traffic‑intelligence capabilities to edge CDN/ENS resources, enabling near‑real‑time video ingestion and unified device management.

Q&A Highlights

YurtHub configuration : Only the cloud‑side access address needs to point to the local YurtHub listener ( http://127.0.0.1:10261); other settings remain unchanged.

OpenYurt vs. KubeEdge : OpenYurt enhances Kubernetes without modifying upstream code, offering better compatibility for native Kubernetes users, whereas KubeEdge introduces deeper changes to kubelet and proxy mechanisms.

Supported OS and kernels : AliOS, CentOS, and Ubuntu with kernel versions ≥ 3.10 are used in large‑scale deployments.

Node scale in examples : Both showcased clusters exceed 100 nodes.

GPU node role : GPUs are used for inference tasks, while model training occurs in the cloud.

Edge autonomy on network loss : Workloads continue running after node or service restarts; full cloud‑side recovery is planned for future releases.

Future edge scenarios (next 3 years) : CDN, edge AI, video, 5G MEC, IoT are already widespread; upcoming domains include vehicular networks, cloud gaming, and traditional industries such as agriculture and energy.

Conclusion

OpenYurt demonstrates how a non‑intrusive, Kubernetes‑native approach can bridge cloud and edge, delivering scalable management, autonomy, and seamless migration while supporting heterogeneous hardware, making it a practical foundation for large‑scale edge deployments.