Key Insights on System Architecture, Evolution, and Feedback Loops

Drawing from work at eBay, Ctrip, and Vipshop, the author presents a platform‑centric view of software architecture, emphasizing that architecture is about identifying stakeholders and addressing their concerns.

Architecture is defined as a system of elements and relationships that satisfy stakeholder needs; each system has an architecture, documented through multiple viewpoints that map to stakeholder concerns.

Effective architects must discover all relevant stakeholders—product managers, developers, testers, ops, etc.—and balance conflicting concerns such as manageability versus performance.

The second core definition highlights that architecture primarily addresses non‑functional requirements (the "‑abilities" such as scalability, reliability, and maintainability).

Architecture should be evolutionary: start with prototypes, perform load testing, then iterate incrementally. The Rational Unified Process (RUP) is cited as a framework that drives architecture with use‑case and risk focus, encouraging iterative development and continuous feedback.

Closed‑loop feedback is essential. The author recommends three feedback paths: system‑level monitoring (network, storage), application‑service monitoring (business logic, CI/CD), and customer‑experience monitoring (user behavior), forming a cycle of collect → measure → adjust → repeat.

Microservices are discussed as an evolution from monolithic designs, stressing single responsibility, independent deployment, and reduced cost of change, while noting the added operational overhead and the importance of team size and complexity.

Conway’s Law is invoked to explain how organizational communication structures shape system designs; misalignment leads to inefficiencies, and aligning architecture with organization is crucial for successful DevOps and platform adoption.

The article also covers the mindset and soft skills of architects: staying open‑minded, valuing relationships over correctness, recognizing the political nature of technical decisions, and continuously improving through hands‑on practice.

Controversial topics such as the relationship between technology and business, and whether architects should write code, are addressed with a nuanced "it depends" stance.

In conclusion, the author urges Chinese architects to keep learning, share knowledge, and aim to produce world‑class architectural leaders.