Private Deployment Practices and Rapid Deployment Platform for JD Transaction Center Microservices

Background : With explosive traffic growth, JD's Transaction Center migrated from monolithic to micro‑service architecture, facing new complexities in resource initialization, multi‑service coordination, and private‑environment deployment.

Private‑deployment Pain Points : Low automation in delivery pipelines, manual image handling, resource provisioning, program verification, and configuration changes made large‑scale deployments chaotic.

Solution Overview : The goal is a one‑time build that can be deployed anywhere via separate image and configuration layers. Configuration is treated as an external, environment‑specific variable that can be templated.

Rapid‑Deployment Platform Architecture (LuoHanTang) : Consists of a Management Center (metadata configuration, deployment orchestration) and a Control Center (deployed inside the customer network to handle middleware registration, resource requests, and deployment execution).

Design Principles :

Standardization – clear rules for image‑configuration decoupling and middleware scope.

Application Self‑Description – applications declare resource and middleware dependencies.

Process Automation – adopt cloud‑native OAM‑style separation of concerns, using environment‑variable templates and automated deployment pipelines.

Metadata Description Language : Defines five component types – Description, Configuration, Resource, Task, and Dependency – allowing developers to describe an app’s resources, middleware, and runtime state without modifying code.

Drag‑and‑Drop Component Design : Components are modeled in three layers (Presentation using JSON Schema, Business Logic for identity‑based execution, and Data for manual/automatic filling). Users compose deployment flows by dragging components, which generate metadata and configuration files that bind resources automatically.

Startup Status Automatic Validation : Implements Kubernetes‑style probes (exec, TCP, HTTP) and additional checks for dependent middleware (e.g., MySQL connectivity) to verify successful service launch.

Results : In a POC, the platform deployed all applications of the Transaction Center within one week, reducing personnel cost by 80% while slightly increasing deployment time, demonstrating significant efficiency gains.

Future Plans : Provide resource‑assessment based pricing, integrate one‑click commercial acceptance testing, and expand middleware support to open‑source and multi‑cloud environments.