want don't You consistency no stinking
summary
Argues for event‑sourced systems that drop strict global consistency to guarantee availability, using git‑like replication and per‑stream versioning to detect divergence by comparing original and current stream positions and emitting a $conflict_detected event. Typical conflicts require two independent writes within ~500ms and are rare; basic detection and notification is implementable in under ~100 lines of code (often ~20). Supports branching as extra streams, scales across clients and nodes (examples: hundreds to ~10,000 partitions), and shifts resolution to humans or simple automation. guarantee availability over global consistency
AI-written summary · about
chapters (5)
- 0:00Start
- 4:04Quorum read' assures consistency!
- 6:30CAP?!
- 26:36Geographic Distribution!!
- 28:13When should you do this?!
related talks
44mDistributed systems and scalability rules
01h 12mSimplifying global-scale strong consistency
01h 01mLost in transaction Strategies to manage consistency in distributed systems
41mConsistency without consensus in production systems
01h 02mAt Least Once
45mDo Distributed Systems they said. It'll be fun they said.
25mThe shell game called eventual consistency
34mAtomicity and Consistency Across Databases and Messaging Systems