Rethinking Reactive Architectures
summary
Argues that reactive architecture is useful but not a silver bullet, showing concrete patterns: event-centric persistence, outbox pattern, listen-to-self, and local event sourcing; maps them to tools like Kafka/Pulsar/Redis, CDC, GraphQL subscriptions and server-sent events for end-to-end push. Recommends projecting aggregates into read models to co-locate data with customer-journey services, cites a mobile-banking finding that 93% of usage is balance/transactions, and describes a sharded production design (24 partitions) that delivered the claimed resilience and throughput.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:43What is reactivity
- 21:10Customer Journeys
- 24:07Who owns data
- 25:06Projecting data
- 27:47Event sourcing
- 30:43What is event sourcing
- 35:03Too long
- 37:28Persistence
- 41:13Decomposition
- 43:21Scaling
- 44:53Sharding
- 49:00Summary
