Evolution of the Application Center Platform with OAM and KubeVela

Background: In recent years cloud‑native platforms have matured, and the concept of cloud‑native applications is being practiced in many companies. The Open Application Model (OAM) proposed by Alibaba and Microsoft, and the CNCF project KubeVela, provide new ideas and frameworks for developing and delivering cloud‑native applications.

Our company launched the development of an Application Center Platform based on KubeVela. After multiple iterations, the platform has been fully deployed internally, supporting hybrid‑cloud delivery, multi‑active governance and decommissioning scenarios, greatly improving application management efficiency and flexibility.

The article describes the evolution of the platform across four stages.

Stage 1 – Building the delivery chain around KubeVela (V1)

Goal: support rapid delivery in hybrid‑cloud environments.

Architecture: four layers – Application Management (definition & delivery), Resource Delivery Orchestration, KubeVela (reusing open‑source components), and resource Operators/Add‑Ons.

Delivery example: user defines an application (Git repo, cluster, MySQL, Redis, monitoring), fills parameters, the platform creates a DAG for delivery, KubeVela invokes Operators, and the user sees the result.

Limitations: only new services can be onboarded; lifecycle coverage is short; unable to automatically detect downstream changes.

Stage 2 – Adapting the platform for existing (stock) data (V2)

Added a synchronization chain (bottom‑up) alongside the delivery chain (top‑down).

Synchronization collects real‑time runtime state of clusters, while delivery handles declarative user intents.

Introduced cluster cloning and resource detection capabilities to automate onboarding of legacy services.

Benefits: improved user experience, continuous data management, reduced manual effort.

Stage 3 – Standardized application management (V3)

Introduced a new “application model” that aligns with OAM concepts but is built from bottom‑up data collection.

Added model management, a custom workflow engine, and offline data synchronization.

Model delivery allows defining model specifications and ensuring consistency across multiple clusters.

Limitations: model functionality is most valuable in multi‑active scenarios; some developers find it redundant for single‑cluster use.

Stage 4 – Multi‑active delivery and governance (V4)

Focus on governance rules (blocking) and delivery rules (advisory) to ensure high‑risk compliance.

Addressed multi‑active resource delivery (e.g., MySQL, Redis) and built productized interfaces for developers.

Implemented cluster cloning, resource detection, and automated model generation to keep data accurate (>95% accuracy).

Explored cost, non‑core dependency, infrastructure, and business‑goal variations when defining norms.

Overall, the Application Center has been iterated for over two years, achieving more than 60% improvement in delivery efficiency through resource detection and cluster cloning. It demonstrates a practical cloud‑native application lifecycle within the company, bridging the gap between OAM theory and real‑world deployment.