Using Multi-Master Data Replication for Refactoring via Parallel Run
summary
A payment system architecture and transactional flow are analyzed, contrasting ideal design with the real-world implementation. Using the example of a surge in UkrPoshta stamp sales, the talk describes problems encountered by this system and other Ukrainian fintechs and banks, and explains step-by-step how multi‑master data replication was used for a parallel‑run refactoring to identify and tune bottlenecks in key components.
AI-written summary · about
chapters (11)
- 0:00Refactoring intro
- 1:02Refactoring reasons
- 3:51Context extraction
- 6:44Parallel run pattern
- 8:51Multi-master replication
- 13:01Data sync methods
- 15:53Solution: pgoutput
- 18:44Consistency layer role
- 21:12New use cases
- 22:38One-year insights
- 26:36Q&A
related talks
15mPlatform Engineering for FinTech
41mAsynchronous Replication Without Censorship
50mReplication based Multi-Master solutions and design patterns
41mAsynchronous Replication Without Censorship
53mMyths and realities of multi-master architecture in PostgreSQL
28mTobias Petry :: Scaling Databases
29mEverything You Ever Wanted to Know About MariaDB Replication but Were Afraid to Ask
28mPractical Approaches To Scaling Distributed Transactions