

The 1000x faster financial database
9 snips Apr 2, 2025
Joran Dirk Greef, the mastermind behind TigerBeetle, a revolutionary distributed database optimized for financial transactions, shares his journey from identifying flaws in traditional databases to developing a solution that achieves astonishing speed. He delves into the architecture that allows for significant improvements over MySQL and Postgres, touches on the advantages of the Zig programming language, and discusses the importance of open-source practices in a competitive landscape. Joran also illuminates the need for tailored solutions in fintech, enhancing transaction efficiency for underserved markets.
AI Snips
Chapters
Transcript
Episode notes
The 76 TPS Limit
- Joran Dirk Greef consulted on a central bank switch project, optimizing its transaction processing.
- Despite hardware upgrades, the MySQL database remained capped at 76 transactions per second.
Row Locks as Bottleneck
- The bottleneck wasn't hardware, but row locks held during SQL transactions across network latency.
- This limits general-purpose databases to around 1000 transactions per second, especially with high contention.
Need for a Transactional Database
- Traditional SQL transactions can handle diverse tasks, but financial transactions have specific needs.
- A dedicated transaction database, optimized for debits and credits, is needed for better performance.