How Does a Matching Engine Work
summary
Explains how a futures exchange matching engine is built as a modular monolith that processes all state changes via a single delivery queue, uses Vertex early then a custom event bus plus Spring Boot for wiring, and guarantees determinism by logging every input to an audit log replay for regression testing (they replay up to a year of real traffic). Throughput targets reach thousands of messages per second and the design yields far fewer regressions and smooth migrations.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 2:21Market evolution
- 4:24Forward contracts
- 6:14Future contracts
- 7:42Modern exchange components
- 10:03Matching engine mechanics
- 12:28Risk and rules
- 14:10Market connectivity
- 16:41Technical challenges
- 20:26High availability
- 22:15Actor-like concurrency
- 24:31Architecture components
- 27:30Auditability and testing
- 30:23Pragmatic high availability
- 34:01Future architecture
- 36:14Scheduler evolution
- 38:41Conclusion
- 41:47Q&A: State recovery
- 44:34Q&A: Technology stack
- 58:52Q&A: Operations and scaling
related talks
17mExtreme Engineering-Elements of Building a High -scale Broking Application
55mFrom the engine to the bridge - building systems that scale
34mDesigning, Building and Testing Deterministic HFT Systems
29mHow to Replace a Jet Engine of your System In Flight
32mChanging Engines Inflight - A Real Life Modernisation Story
56mProcessing Real Time Trading Data with Event Streaming
40mCommunication between (micro-)services
01h 06mPatterns for scalability and availability in (trading) systems