EventStoreDb: To Be or Not to Be, That Is the Question
summary
This talk describes running a large .NET event-sourced system using EventStoreDB (C#) to persist workflows for ~300 million entities/month and 1–2 billion events/month. It covers real production problems—disk I/O saturation, a bug in persistence subscriptions that missed reads, too many open files, and costly deletions—and the mitigations used: switching JSON to protobuf for smaller payloads, resizing clusters and instances, and other fixes to restore throughput and reliability.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 7:05State storage problem
- 14:18EventStoreDB overview
- 21:09Technical challenges
- 26:33Results and plans
- 31:33Cluster reliability Q&A
- 33:43Data cleanup
- 36:57AWS comparison
- 38:04In-memory storage design
- 40:58Q&A
related talks
48mEvent Sourcing - what could possibly go wrong?
01h 08mEvent Driven Architecture – Enabling Microservices
57mEventuous Event Sourcing with .NET
52mEvent Sourcing for the Cloud Developer
48mEvent-Driven Architecture – Benefits and Pitfalls
45mReliable Integration Architecture with EventStoreDB
36mThe Good, The Bad and The Ugly - An Event Sourcing Retrospective
43mUnraveling the Threads: Event-Sourcing, CQRS, and Event-Driven Architecture