Dealing with eventual consistency
summary
Shows a demo movie-ticket site that uses SignalR acknowledgements and an Azure Service Bus queue to accept orders immediately while backend processing is asynchronous; localStorage provides an instant optimistic UI so comments and subscriptions appear without waiting. Explains single-item hotspots (Game of Thrones, tens of thousands of simultaneous requests), shard/index delays, and a lottery pattern that safely accepts 50k registrations then allocates tickets later. Includes a working demo where backend shutdown leaves messages on the queue and SignalR later reports final results.
AI-written summary · about
chapters (16)
- 0:00Eventual consistency intro
- 3:57High traffic struggles
- 6:04Transactional challenges
- 9:15Relational scaling issues
- 12:07Messaging for consistency
- 14:17Kitten analogy explained
- 17:26User expectations during delays
- 21:30Movie ticket demo
- 27:28SignalR for real-time
- 31:00GoT lottery system
- 36:21Technical solution
- 37:26Event sourcing for congestion
- 41:07When accuracy matters and when it doesn't
- 45:33Business concurrency fixes
- 49:08Eventual consistency is not scary
- 50:12Conclusion & Q&A
related talks
39mDealing with eventual consistency
58mWelcome to the (state) machine
49mBuilding Event Driven User Interfaces
47mPatterns for Building Event-driven Web & Mobile App Backends
38mBeyond REST: Using Full-Stack Signals for Real-Time Reactive UIs
28mRace Conditions in JavaScript Apps
42mReactive applications with WebSockets
54mWhere we're going... we don't need batch jobs