From RDBMS to NoSQL (PRT314)
summary
Amazon runs a project called Rolling Stone to replace Oracle across ~3,000 servers and ~10,000 services supporting 25–30k developers; analysis shows ~70% of access patterns hit a single row and ~20% a range, while ~10% of JOIN-heavy queries drive half the infrastructure cost. The practical fix is document modeling: use multi-key indexes, key-prefixing and an outlier (bucket) pattern, measure with Schema Bench, and favor O(log N) indexed lookups — DynamoDB proof reduced cart costs ~80% and MongoDB’s storage/compression (WiredTiger ~16 KB block) limits write-unit waste.
AI-written summary · about
chapters (12)
- 0:00Introduction to NoSQL
- 3:31The case for NoSQL at Amazon
- 6:19History of database technology
- 9:19The limitations of SQL
- 13:07Data modeling relationships
- 17:39Document model efficiency
- 25:33Overcoming adoption hurdles
- 30:19Developer enablement tools
- 33:37NoSQL methodology
- 36:52Use case: single view
- 40:56Advanced design patterns
- 48:22Conclusions and takeaways
related talks
59mModernize your applications with purpose-built databases
41mDo you really need that relational DB?
01h 00mFrom Relational Databases to NoSQL
29mPurpose-Built Databases for Modern Applications
21mUse cases that drive the need for NoSQL databases
29mPurpose-Built Databases for Modern Applications
18mFrom SQL to NOSQL - a gentle intro for Devs
40mBeyond Relational: Data storage for modern applications