How Tencent Builds a Hybrid Cloud‑Native Industrial IoT Platform with Edge Containers

1. Demand Analysis

Tencent identified three core demands for the industrial internet: security (especially data protection), empowerment (leveraging massive cloud compute to serve the entire manufacturing value chain), and ecosystem (building a collaborative platform that connects partners, operators and customers).

2. Architecture and Modules

Overall Architecture

The platform adopts a three‑layer hybrid structure:

Cloud layer : Tencent Cloud provides the full suite of compute, AI and big‑data services.

Regional/Industrial cloud layer : A limited‑capacity cloud (industrial cloud) is deployed in specific regions to deliver AI inference, IoT backend, alerting and customized operational policies.

Enterprise data‑center layer : Sensitive data and workloads run in the customer’s own data center, forming a true hybrid cloud.

An open platform sits on top to integrate ecosystem partners, while operational cost reduction is a key design goal.

Industrial Cloud Design

The industrial cloud consists of a control side hosted on Tencent Cloud and an edge side located in the nearest regional data center. The control side connects to the edge side via Tencent Cloud services.

Key modules include:

Product Center : Bridges Tencent Cloud Marketplace SaaS and AI applications to the industrial cloud.

Sync Center : Enables bidirectional data synchronization between control and edge nodes.

Message Center : Provides global routing and notification across isolated industrial clouds.

Industrial Cloud Core Modules

Important components are:

Account Center : Uses shadow accounts on Tencent Cloud to enable seamless hybrid operations.

Billing Center : Supports diverse billing models for public, private and hybrid clouds.

Ti‑EMS : AI‑elastic cluster service for training in the public cloud and inference in the industrial cloud or edge nodes.

Infrastructure : IaaS (TStack, ECM, TCE) and PaaS (MySQL, Redis, MQ, Object Storage) layers provide foundational capabilities.

Supporting tools include Go‑based development framework, Coding CI, TKE CD, Prometheus monitoring, CLS logging, and Flyway database migration.

3. Edge Containers

Edge containers combine a cloud‑side Kubernetes Master with an edge‑side Kubernetes Node , linked via a tunnel. This design ensures data locality, low latency for AI quality inspection, and cost savings by minimizing edge‑side components.

Key technical points:

Tunnel Edge : Establishes a secure tunnel between edge nodes and the cloud master.

Lite‑apiserver : Acts as a gateway with edge‑autonomy, keeping pods running when the public network is disrupted.

Observer : A health‑checking module that monitors node liveness and coordinates pod migration in case of failures.

Operations shift from manual VPN/SSH upgrades to a DevOps workflow where authorized users trigger version upgrades from the cloud console, and Kubernetes performs rolling updates automatically.

4. Platform Products and Solutions

AI Quality‑Inspection Platform

A manufacturing customer uses cameras to capture LCD panels, annotates defects, and trains a model on Tencent Cloud Ti‑ONE. The trained model is deployed to edge nodes for real‑time inspection, eliminating manual visual checks.

The workflow includes a training pipeline (weeks of compute on Ti‑ONE), inference service on Ti‑EMS, and a scheduling system for multi‑line production orchestration.

IoT Industrial Equipment Monitoring

Sensors collect device metrics (temperature, vibration) and send them via gateways to an IoT backend, which can run on the industrial cloud or edge nodes. A rule engine and alert system, backed by Tencent’s big‑data suite (TBDS), enable predictive maintenance and anomaly detection.

Other Capabilities

The platform also offers low‑code cloud development, collaborative manufacturing tools for regional supply‑chain partners, a product‑traceability (code‑parsing) service, and integration with third‑party SaaS applications through open APIs.

5. Q&A Highlights

Edge nodes are primarily deployed at the city level, with some provincial or industrial‑park deployments. The solution favors SaaS over container images for manufacturing customers, but both are supported. The open platform allows third‑party SaaS to integrate with account, billing and UI standards.