SuperEdge

SuperEdge is an open-source container management system that extends native Kubernetes to edge computing environments without modifying the Kubernetes source code. It manages compute resources and containerized applications across multiple geographically distributed edge regions, treating them as a single Kubernetes cluster.

Developed by Tencent Cloud and co-launched with Intel, VMware, Huya, Cambricon, Captialonline, and Meituan, SuperEdge became a CNCF sandbox project. It targets scenarios where edge nodes have unreliable or intermittent connectivity to a central cloud control plane — power substations, highway management systems, retail sites, and distributed industrial facilities.

Key features

• Edge autonomy (L3/L4/L5): At L3, nodes continue running existing workloads when disconnected from the cloud. At L4/L5, the Kins feature provisions an embedded lightweight K3s cluster on edge NodeUnits, enabling fully offline operation including write operations.
• Non-intrusive Kubernetes extension: SuperEdge does not replace kubelet or modify Kubernetes internals. All standard Kubernetes APIs and resources remain compatible, and existing clusters can be converted to SuperEdge clusters with minimal changes.
• Distributed health monitoring: Edge-side health checks (edge-health daemon) run peer-to-peer across nodes in the same edge region, providing more accurate fault detection than centralized cloud checks alone.
• ServiceGroup: A custom CRD workload unit (DeploymentGrid, StatefulSetGrid, ServiceGrid) that closes network traffic within a geographic region. Canary deployments are natively supported.
• Network tunneling: The tunnel-cloud and tunnel-edge components maintain a persistent reverse proxy connection supporting TCP, HTTP, HTTPS, and SSH, allowing the cloud control plane to reach edge nodes behind NAT or firewalls.
• WasmEdge support: SuperEdge supports containerd + crun + WasmEdge for running WebAssembly workloads at the edge, enabling lighter-weight container images and faster startup times.
• edgeadm installer: A kubeadm-based CLI for bootstrapping SuperEdge clusters, supporting amd64 and arm64 architectures.

Architecture

SuperEdge uses a split cloud/edge component model. Cloud-side components (tunnel-cloud, application-grid controller, edge-health admission, site-manager) run in the central Kubernetes master. Edge-side components (tunnel-edge, lite-apiserver, edge-health, application-grid wrapper) run on each edge node. The lite-apiserver acts as a local cache and proxy for the Kubernetes API, enabling nodes to serve requests during disconnection.

Getting started

Deploying SuperEdge requires edgeadm, which wraps kubeadm. Supports amd64 and arm64 architectures; tested against Kubernetes 1.20.6 and 1.22.6 (as of v0.9.0).

arch=amd64 version=v0.9.0 kubernetesVersion=1.22.6
wget https://superedge-1253687700.cos.ap-guangzhou.myqcloud.com/$version/$arch/edgeadm-linux-$arch-$version-k8s-$kubernetesVersion.tgz
tar -xzvf edgeadm-linux-* && cd edgeadm-linux-$arch-$version-k8s-$kubernetesVersion
./edgeadm init --kubernetes-version=1.22.6 --enable-edge=true --edge-version=0.9.0

Limitations

• Requires running kubelet and five edge-side components per node — significantly heavier than KubeEdge's single-component footprint, making it unsuitable for very resource-constrained devices with less than ~1 GB RAM.
• Last release (v0.9.0) targets Kubernetes 1.22.6; the project has not published releases supporting Kubernetes 1.24 or later, creating a meaningful compatibility gap with current Kubernetes distributions.
• Community activity has declined since 2023 — last GitHub commit was February 2024, with 40 open issues and no recent releases, indicating the project may be in maintenance-only mode.
• CNCF sandbox status only (not incubating), meaning it has not passed CNCF's production-readiness due diligence; KubeEdge holds incubating status with a substantially larger community.
• ServiceGroup closed-loop networking prevents edge services from easily communicating across NodeUnits without custom routing, which limits multi-region data aggregation architectures.