Why Tencent Cloud Chose Kubernetes and Its Cloud‑Native PaaS Architecture

Kubernetes has become a hot technology, and many internet companies have built proprietary PaaS platforms on it. Tencent Cloud decided in late 2016 to develop a container product, selected Kubernetes after evaluating alternatives, and launched the managed Kubernetes service (CCS) in early 2017.

During 2017 the team continuously added features such as monitoring, log processing, and container registry, managing thousands of clusters and hundreds of thousands of containers.

Why did Tencent Cloud choose Kubernetes over other orchestration tools? The article lists the core ideas of Kubernetes, Mesos, and Docker Swarm, then explains that Kubernetes inherits design from Google’s Borg, has a large active community, supports stateful services, and is backed by the CNCF.

Key advantages cited include: large ecosystem, strong community support, ability to run stateful workloads, and suitability for large‑scale micro‑service deployments.

Tencent Cloud now offers a suite of container solutions, including the managed Kubernetes service, single‑container instance service, TencentHub, Flow Engine, and CCI, as well as Kubernetes‑based TensorFlow and Spark solutions.

The managed Kubernetes service provides full compatibility with native Kubernetes APIs, integrates with Tencent Cloud plugins (CBS, CLB), and supports configuration management, application templates, and service groups.

Configuration management is realized through Helm templates, ConfigMap environment variables, and ConfigMap volumes, enabling versioned configuration, easy roll‑back, and multi‑environment deployments.

Application templates describe multiple services with default configurations, allowing rapid cloning, multi‑environment deployment, and marketplace‑driven template consumption.

For high‑availability, Tencent Cloud runs the master components of user clusters as containers inside a dedicated Kubernetes cluster, sharing etcd and control‑plane nodes, eliminating the need for dedicated VMs or agents, and simplifying upgrades and roll‑backs.

The high‑availability architecture separates the control plane (master components and service‑management nodes) from the data plane (user workloads). The control plane consists of a highly available set of master nodes and an etcd cluster, while user nodes run workloads.

Future plans include expanding container runtimes (Docker, OCI, Kata), more DevOps services, and public beta releases of single‑container instance service and TencentHub.