42 Lessons Learned from Building a Production Database (Translation)

Customer (User) – Keep customers happy, start with a single customer, engage directly with them, understand their real needs by reading their code, and recognize that customers may not always articulate their true requirements.

Project Management – Define a clear mission statement, repeatedly assess task difficulty, allocate tasks to individual contributors (ICs) who are on the critical path, treat road‑maps as means, ensure managerial changes do not unfairly affect ICs, and benchmark task durations against similar projects.

Design – Be conservative with APIs and liberal with implementations, introduce new implementations gradually, design APIs with a single CLI switch, prioritize consistency and durability over availability for storage systems, maintain multiple implementations for testing, use late binding for designs and implementations, manage abstraction levels, avoid relying on real‑time clocks for correctness, and foster a culture of continuous design exploration.

Code Review – Maintain a transparent, fast‑review codebase, encourage critical diff analysis, require multiple approvals for critical components, resist rushing implementations, and create an environment where discarding flawed code is acceptable.

Strategy – Regularly question the team’s purpose, track related projects to understand design trade‑offs, compete on fundamental design characteristics rather than raw performance, and be willing to adopt superior external solutions.

Observability – Treat measurement as a means, detect issues before customers do, place observability above APIs and implementations, pay attention to hard‑to‑measure properties like consistency, and push critical checks into deployment rather than external services.

Research – Keep up with domain research, experiment with new ideas, avoid copying designs verbatim, and write papers to clarify assumptions and attract talent.