Evolution and Platformization of the Auto Home E‑commerce System Architecture

Preface – The Auto Home e‑commerce system was launched in 2014, grew rapidly between 2016‑2019, and survived multiple peak sales events, establishing a reliable, high‑performance online transaction capability.

Architecture Evolution

1. Initial Stage (2014‑2015) : Rapid iteration required a shift from .NET to Java, culminating in a full migration to a Java stack and the launch of a complete online car‑shopping platform.

2. Micro‑service Stage (2016‑2018) : Scaling pressures led to splitting a monolithic front‑end project into dozens of Maven modules, causing merge conflicts and performance issues. The solution was a distributed architecture with independent centers for users, products, orders, promotions, coupons, merchants, etc., improving development efficiency and system stability.

3. Master‑Data Stage (2019) : To support diverse transaction models, a unified master‑data layer was built, standardizing data models for merchants, products, orders, and promotions, and synchronizing heterogeneous data via binlog ingestion.

4. Platform‑centric Stage (2020‑present) : The platform focuses on business identity, service orchestration, and extensibility to reduce coupling, enable rapid feature assembly, and support multiple business lines.

Platform‑centric Architecture Practice

The platform separates business identity, service orchestration, and configuration. Business identity (person, product, venue) uniquely identifies each transaction. Service orchestration combines micro‑services into reusable business capabilities, while extension points (SPI/COLA) allow plug‑in implementations for specific scenarios.

Key layers (bottom‑up): infrastructure, basic services, business capability, business process (orchestration), and business layer (identity, extensions, workflow).

Business Identity – Modeled after Alibaba’s tenant concept, identities are defined across three dimensions (person, product, venue) to isolate data and route services.

Service Orchestration – Micro‑services are composed via orchestration frameworks (e.g., Apache Camel) rather than hard‑coded calls, enabling visual workflow design and reusable component assembly.

Extension‑Point Framework – Uses Java SPI and Alibaba’s COLA framework to decouple interface from implementation, supporting multi‑level routing (useCase, bizId, scenario) for flexible business logic.

Practical Examples

Order‑cancellation logic was refactored from hard‑coded sequences to a configurable orchestration with extension points, reducing response time by ~30% and cutting regression testing effort.

Business Configuration – Core transaction parameters (e.g., auto‑push to resource pool, ID‑card requirement, auto‑confirm receipt) are managed centrally, allowing per‑business variation without code changes.

Development Tooling – A suite of internal tools (message distribution, price calculators, cache management, one‑click degradation) improves developer productivity.

Data Visualization – Real‑time and offline dashboards monitor order volume, resource usage, and trigger alerts via DingTalk when thresholds are exceeded.

Knowledge Accumulation – Technical and product experience is captured in an internal knowledge base with searchable APIs, fostering continuous learning.

Conclusion – The platformized architecture has enabled the Auto Home e‑commerce system to support diverse business models, handle large‑scale sales events, and lay groundwork for future innovations such as new‑retail and further platform upgrades.