Vinted's engineering team shares their experience building a globally distributed, event-driven platform that handles 300k+ requests per second. They chose a single-writer, multi-reader architecture with centralized writes and replicated read projections across regions to balance consistency with performance. The approach involves designing for eventual consistency, building read-optimized data projections from event streams, and fundamentally shifting team culture to embrace distributed systems thinking including retries, idempotency, and partial failures. While the foundation is in place, ongoing work focuses on strengthening async tooling, improving developer experience, and scaling infrastructure to support continued global growth.
Table of contents
Designing for a Global FootprintFaster Reads Through Data ProjectionsThe Cultural Shift Behind the ArchitectureWhere We Are, and What’s NextWhy This Work Matters, and Why You Might Want to JoinSort: