DYNAMO: the paper that changed the database world
summary
Explains the Dynamo paper claim that a masterless key value store can run active active over multiple datacenters and support millions of durable writes a second with single digit millisecond response times. Shows concrete mechanisms used: consistent hashing, replication strategies, Gossip for membership, sloppy quorums, hinted handoff and anti-entropy repair, and how Apache Cassandra implements or alters those choices and where conflicts and trade offs appear in real deployments.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 1:55Takeaways
- 4:20Axis of a DB
- 8:11Single node DB
- 8:31How do we scale databases?
- 8:44Horizontal scaling
- 8:55Dynamo databases
- 9:19CAP
- 11:16PAC-ELC
- 12:39Master slave
- 13:13Scaling reads
- 13:36Scaling writes
- 14:27Getting to Dynamo
- 14:45Consistent Hashing
- 17:37CH: Master?
- 18:32CH: Quorums
- 21:15CH: Sloppy Quorums
- 24:29Conflict: the data model
- 25:28Conflict: Clock time
- 29:02Conflict: Vector Clocks
- 33:12Conflict: Immutable
- 39:20Summary
related talks
27mDissecting Dynamo: Highly available key value store
41mEverything you always wanted to know about highly available distributed databases
57mLiving in eventually consistent reality
37mDynamoDB in an EDA World
33mDynamoDB Essentials and Integration for Java Developers
30mBuild for Scale, Performance With Feature-packed Amazon DynamoDB
57mScaling the CAP Theorem for Modern Distributed Systems
20mAmazon DynamoDB: A Scalable, Predictably Performant, and Fully Managed NoSQL...