Scaling Vinted: A System That Never Stands Still

on YouTube ↗37m

" summary

Vinted documents moving from a Rails monolith on MySQL to sharded MySQL with Vitess, then modularized in-place and split into coarse-grained services with enforced functional sharding and strict data ownership. Event-driven projections run through Kafka with an event store built on Apache Flink and Iceberg to seed local databases, enabling edge read sites that already handle ~10–15% of peak traffic on five machines. Ongoing work tackles distributed writes, consistency tradeoffs, and new storage needs.

AI-written summary · about

" chapters (18)

1. 0:00🚀 Build Stuff Conference – Get Your Tickets!
2. 0:22Introduction: Scaling Vinted From App to Distributed Marketplace
3. 2:10Vinted’s Mission, Business Scope, and Engineering Scale
4. 4:15Core Foundations: On-Prem Infrastructure and Technical Control
5. 6:20Ruby on Rails, MySQL, and Adopting Vitess for Database Scaling
6. 8:35Continuous Delivery, SRE Mindset, and Reliability as a First-Class Citizen
7. 10:45Rapid Growth Across Europe and the Limits of the Shared Codebase
8. 13:20Why Hiring More Engineers Didn’t Solve the Architecture Bottleneck
9. 15:05Breaking Up the Monolith: Modularization as the First Step
10. 18:10Facades, Rails Modules, and Architectural Fitness Functions
11. 20:50Early Wins, Developer Friction, and Hidden Complexity in the Monolith
12. 23:15Rising Traffic Peaks, Bigger Incidents, and the Need for Service Isolation
13. 25:05Moving to Coarse-Grained Services Without Falling Into Microservice Chaos
14. 27:10The Real Bottleneck: Shared Database Coupling and the Cost of Slow Queries
15. 28:50Functional Shards, Data Ownership, and Database Boundary Enforcement
16. 30:40Preparing for Global Scale: Edge Sites, Latency, and Speed of Light Constraints
17. 32:20API Gateway Design, Global Routing, and Read-Only Edge Architecture
18. 34:00Future Challenges: Distributed Writes, Consistency Tradeoffs, and New Storage Needs

" related talks

• 46mScaling YouTube's Backend: The Vitess Trade-offsSugu Sougoumarane · @Scale 2014
• 37mSharding: Growing Systems from Node-scale to Planet-scaleAdam Mckaig · SREcon24 Americas
• 52mTaking Storage for a RideRené Schmidt · GOTO Copenhagen 2016
• 50mImmutable Values in the Big Wide WorldTim Ewald, Paul deGrandis · Clojure/conj 2024
• 02h 18mDistributed Systems: In the Daily Life of the Data EngineerFrançois Sarradin · Devoxx France 2024
• 34mShared Nothing Architecture - How to Scale to InfinitySam Ruby · Carolina Code Conference 2024
• 32mData @Scale Seattle- Eric Hwang introduced by Ganapathy KrishnamoorthyEric Hwang, Ganapathy Krishnamoorthy · Data @Scale Seattle 2015
• 51mScaling yout Architecture with Services and EventsRandy Shoup · GeeCON 2019