Experience Developing, Debugging, and Deploying a Hot-Swap Trading System
summary
Presentation explains why and when a hot-standby (active-passive) backup for a trading system is needed, and why Moscow Exchange chose to implement one. It reviews classical hot-standby architectures and prior work, then describes motivations for developing a custom failover algorithm to keep the exchange maximally available even when automated decisions might otherwise halt trading.
AI-written summary · about
chapters (11)
- 0:00Hot-swap trading intro
- 1:27High-frequency growth challenge
- 3:53System architecture & design
- 8:34Debugging rounding bug
- 12:24Adding warm standby servers
- 15:35Consensus hot-swap system
- 20:20Consensus algorithm workflow
- 25:21Server recovery handling
- 29:17Testing challenges
- 36:30Key lessons & limits
- 42:33Q&A
related talks
11mMitigating Against Large Scale Systemic Failures in E-Trading
01h 06mPatterns for scalability and availability in (trading) systems
50mEvolution of the Moscow Exchange Trading and Clearing System Architecture
42mStandby in Combat: Scaling an App in the World's No.2 Classifieds
34mDesigning, Building and Testing Deterministic HFT Systems
54mDistributed Caching in Real-Time Payment Systems
47mProduction Engineering When Trading Billions of Dollars a Day
25mSmoke & Mirrors: The Primitives of High Availability