Ant Group Data Architecture Practice and Upgrade Strategy

The article begins with an overview of data architecture concepts, distinguishing technical architecture from content architecture and emphasizing the need for a clear, macro‑level structure to manage growing system complexity.

It then analyzes Ant Group's existing data architecture, highlighting issues such as resource contention, unclear domain boundaries, and tangled dependencies caused by a massive graph of tables and pipelines.

Three common architectural patterns are examined: the silo (chimney) model, the large‑mid‑platform model derived from Alibaba's OneData approach, and the domain‑driven (Data Mesh) model, each with its advantages and drawbacks.

Based on these findings, a new upgrade plan is proposed that reorganizes concepts around a unified "domain" hierarchy, introduces clear responsibilities, resource isolation, and modular development, and applies domain‑driven design to define data applications, units, and code repositories.

The implementation guidelines include single‑responsibility, accountability, and resource‑isolation principles, as well as detailed governance layers (company‑level, domain‑level, and application‑level) to ensure consistent standards.

In the outlook section, the data system is modeled as a complex, scale‑free network, revealing properties such as robustness to random failures but vulnerability to targeted attacks, low propagation thresholds, and high critical immunization rates.

The article concludes with recommendations to protect hub nodes, use network simulations for architectural impact assessment, and adopt a complexity‑oriented mindset for managing the evolving data ecosystem.