Consistent Updates in a Distributed System without Transactions
summary
Concrete patterns for making distributed updates consistent without transactions using idempotent post-processing events and optimistic logging for cross-service handoffs. Covers capturing one-way events to drive eventual state, end-to-end failure and retry handling including how postprocessors detect and repair inconsistencies, and deployment notes using Google Cloud Platform with Cassandra and relational databases; design choices are motivated by production numbers (e.g., a basket with 4,000 items and systems creating 422 store snapshots).
AI-written summary · about
chapters (10)
- 0:00Consistent updates in distributed systems
- 3:07Service calls in microservices
- 8:42Response time and QoS
- 15:07Conflict resolution in data models
- 19:11Post-processing for data views
- 22:17Live data capture and events
- 25:34Network latency and timing
- 27:33Production insights from events
- 31:23Large-scale distributed case study
- 33:52E-commerce basket challenges
related talks
39mManaging Consistency, State and Identity in Distributed Microservices
01h 02mConsistency and Agreements in Distributed Systems
27mData Consistency in Microservices Architecture
30mData Consistency Patterns in Cloud Native Applications
46mLost in transaction? Strategies to deal with (in-)consistency in distributed systems
25mConsistent Updates for Scalable Microservices
36mLost In Transaction
30mDistributed system design