How to Read the Exact Time When You Have a Billion Clocks
summary
Explains how Conflict-free Replicated Data Types (CRDTs) enable replicas to converge without coordination despite lost, duplicated, or out‑of‑order messages, enabling cheaper large‑scale synchronization. Describes the required math (states as a join‑semilattice; updates that are commutative, associative, idempotent), an in‑house CRDT implementation across three languages, property‑based testing used for verification, and deployment to roughly one billion devices per month to keep edge state consistent.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 5:13Distributed problem
- 9:22Counting events
- 10:03CRDTs and properties
- 15:05CRDT math rules
- 20:43Technical details
- 22:07CRDTs in production
- 23:50Real-world CRDT apps
- 24:12Q&A
related talks
57mLiving in eventually consistent reality
35mData Types for Eventually Consistent Databases
50mCRDTs in Practice
49mDistributed Data: Conflict-free Data Structures in JS
22mHappily Ever After: A CRDT Fairy Tale
46mCoordination-Free Computations
46mPractical data synchronization with CRDTs
21mCRDTs for Mortals