Designing, Building and Testing Deterministic HFT Systems
summary
Chronicles Java libraries and microservices focus on minimal work per component to hit tight SLAs using JMH benchmarks, kernel bypass NICs, and avoiding DB calls. A three‑microservice fixed income OMS achieves 99.9% <20µs end-to-end for six persisted messages, two-process persist and deserialize runs under 1µs, replication adds ~30µs, and a bank replacement yields ROI in three months. Design uses event-sourced deterministic lambdas, low GC churn, long-encoded IDs, data-driven replayable tests and fast component constructor-only wiring for reproducible debugging and failover.
AI-written summary · about
chapters (8)
- 0:00Introduction to deterministic HFT
- 3:36Designing for minimal latency
- 10:07Sources of latency in systems
- 14:52Garbage collection and JVM tuning
- 17:00Deterministic functions and event sourcing
- 22:29Storing everything for reproducibility
- 24:38Unit testing with deterministic inputs
- 32:45Balancing speed and developer efficiency
related talks
39mTrust Deterministic Execution to Scale & Simplify Your Systems
01h 06mPatterns for scalability and availability in (trading) systems
31mDesigning a mission critical event system
43mFault tolerance – look, it’s possible!
51mBuilding Fault Tolerant Microservices
01h 17mWhen Less Is More: Designing & Decoding Complex Distributed Systems
41mTracing Real-Time Distributed Systems
40mCAPtivating architecture: Navigating Distributed Systems and Microservices