Reactive Systems: 21st Architecture for 21st Century Systems
summary
A low-latency trading exchange is built from scratch using ring buffer journaling, clustered durable queues and single-threaded services that handle hundreds of thousands of messages per second with microsecond latencies. The design enforces idempotent messaging, share‑nothing isolation, location transparency, sharding for linear scalability, and back‑pressure to avoid queue overload. Eventual consistency is accepted inside bounded contexts; the system is deterministic, replayable from production messages, and proves to be resilient and highly scalable.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 1:48Our World Is Changing
- 3:57The Reactive Manifesto
- 5:10Reactive Systems Are
- 7:31Properties of Reactive Systems
- 8:26Fractal Architecture
- 11:16Failure Modes in Synchronous Messaging
- 13:30Synch Messaging Breeds Complexity
- 15:51The Benefits of Asynchrony
- 19:32An Example of Idempotence
- 22:48Isolation
- 24:35Share Nothing
- 26:08Queues Represent an Unstable State
- 27:47Back-Pressure
- 28:37Eventual Consistency
- 30:47Location Transparency
- 32:58Linear Scalability Through Sharding
- 34:03Modern Hardware Should Change Our Assumptions
- 36:23Conway's Law
- 37:25Bounded Contexts
- 40:31Example Reactive, MicroService architecture
- 45:11Where to start?
