The InfoQ Podcast Event Sourcing: Bernd Rücker on Architecting for Scale
Sep 13, 2019
Bernd Rücker, co-founder and chief technologist at Camunda, shares his expertise in workflow automation and event-driven architectures. He delves into the advantages of event sourcing for scalability and contrasts it with traditional RDBMS limitations. The discussion highlights the confusion between commands and events, the importance of sagas in maintaining consistency, and the need for effective orchestration in microservices. Bernd also touches on Zeebe's architecture, emphasizing its strong state guarantees and the role of event streaming for enhanced performance.
AI Snips
Chapters
Transcript
Episode notes
RDBMS Hits Throughput Limits
- The main limitation of an RDBMS for orchestration is throughput and extreme-scale transaction cost.
- Event-based architectures avoid heavy DB sharding and achieve much higher scalable throughput for demanding use cases.
Avoid Distributed Transactions
- Avoid distributed transactions across microservices; use workflows or sagas to manage multi-step consistency.
- Design compensating actions and mini-workflows instead of relying on global transactions.
Use Projections And Exported Read Stores
- Build projections and separate read-models (CQRS) to restore queryability when runtime state is event-sourced.
- Export events to systems like Elasticsearch for operational tooling and accept eventual consistency for reads.