Do you really need that relational DB?
summary
Argues that user‑facing pages can skip relational queries by generating and serving purpose‑built read models (HTML or JSON) on save, using events, queues (RabbitMQ/Kafka) and projectors to produce them; cites a real incident where prebuilt HTML brought a news site back up and notes typical cache hit ratios around 60% and shops with 40–50k products as candidates. Recommends NoSQL with DynamoDB/MongoDB change streams for source of truth, warns about regeneration cost, monitoring and eventual consistency.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 5:25Cache
- 11:27Decouple
- 14:23Legacy
- 17:45Projector
- 25:48product page
- 27:18duplicate data
- 31:22injection points
- 36:09Questions
related talks
40mBeyond Relational: Data storage for modern applications
34mData Storage: NoSQL Toasters and a Cloud of Kitchen Sinks
45mNoSQL: Death to Relational Databases
01h 00mFrom Relational Databases to NoSQL
53mFrom RDBMS to NoSQL (PRT314)
19mRelational or Non-Relational? Choosing the Best Database for the Task
38mDubious Database Design
52mData storage for modern applications - Mike Lehan