Don’t Drop ACID - Transactions in Distributed NoSQL
summary
Argues that many distributed NoSQL systems now support multi-document ACID behavior and shows a working Java demo against Couchbase that commits, rolls back on exception, and leaves staged transaction metadata for a background cooperative cleanup to remove; uses the Couchbase Java SDK on a local cluster, observes a 60s cleanup timeout, and demonstrates both happy-path commit and rollback. Compares client-side vs server-side transaction designs and advises using transactions only when necessary.
AI-written summary · about
chapters (18)
- 0:00Start
- 3:22Who am I? Where am I?
- 5:03Third Normal Form
- 6:25Why Transactions?
- 9:00JSON Document Database
- 12:07NoSQL: The Big Four
- 17:53A is for Atomicity
- 18:32C is for Consistency
- 23:40I is for Isolation
- 24:52D is for Durability
- 27:35"Split Brain" (aka network problems)
- 30:23Latency
- 31:47Correctness
- 33:54Client-side vs Server-side
- 49:37What are the tradeoffs?
- 51:33Other NoSQL ACID Transactions efforts
- 56:13Go down the rabbit hole...
- 57:07Contact me!
related talks
01mDon't Drop ACID - Transactions in Distributed NoSQL
47mBeing transactional in 2019 - what does this actually mean?
52mTransactions: Myths, Surprises and Opportunities
34mCassandra on ACID: This Changes Everything
42mMyths of NoSQL
31mLost in Transaction?
01h 00mDistributed transactions implementation trade-offs
28mPractical Approaches To Scaling Distributed Transactions