CREATE INDEX… USING VODKA. VODKA CONNECTING INDEXES!
summary
Prototype index access method called Vodka appears, replacing GIN’s B-tree entries with an ART/digital tree and combining multiple index types (including R-tree variants) to index full JSONB paths losslessly. JSQuery packs JSON queries into a single value for indexable search; tests on ~1M Delicious bookmarks show JSONB storage overhead <4% and Gin-based indexes far smaller than comparable MongoDB indexes (44 MB vs ~387 MB). A simple optimizer and hints yield >10x speedups; WAL-generalization would let Vodka ship as an extension. JSONB JSQuery Vodka prototype
AI-written summary · about
chapters (12)
- 0:00Introduction to Vodka
- 1:00Why Vodka was created
- 2:24JSONB and its evolution
- 9:29JSQuery extension explained
- 13:06JSONB indexing challenges
- 17:51Postgres vs MongoDB comparison
- 24:41Limitations of JSQuery
- 26:48Introducing Vodka access method
- 30:50Vodka combining access methods
- 34:15Future of JSONB and Vodka
- 35:51Vodka as a Postgres extension
- 37:17Upcoming Postgres conference
related talks
53mAdvanced Indexing Techniques in PostgreSQL
35mLeveraging PostgreSQL 17's JSONB for Enhanced Data Modeling
25mJSON in PostgreSQL - evil data type or just needs to be tamed?
45mSQL Indexing Beyond Simple Columns: Expressions, Text Fragments, Attributes J...
42mNoSQL as Not Only SQL
14mJSON data structure – the hidden gem inside your database engine
48mB-tree Indexes in Databases Illustrated by IndexedDB and PostgreSQL
48mPostgres and Json - How to get the best of them?