Breakout Session
summary
Claims that Postgres matches non‑relational analytics without sacrificing ACID, and demonstrates it by installing and composing real extensions: MADlib logistic_regression_train for supervised models, pg_similarity + MADlib record linkage to join messy datasets, and approximate HLL cardinality (1–2% error) to avoid expensive COUNT(DISTINCT). Uses PGXN, multicorn FDWs, ZomboDB/Elasticsearch indexing and shows MADlib running on MPP Greenplum; an experimental background worker auto‑fills predicted columns, proving the approach works end to end.
AI-written summary · about
chapters (12)
- 0:00Intro
- 2:20The Prevailing View - Logical
- 6:02The Practical Reality
- 8:52Nearness Easy
- 11:40The PostgreSQL Extension Network
- 16:14UDAs & Data Types: postgresql-hll
- 19:49Foreign Data Wrappers: API
- 21:36Indexes: ZomboDB
- 23:48Composing Extension Methods: MADlib
- 29:10Parallel Processing
- 30:30Composing Extensions: Alps
- 34:20Conclusion
related talks
34mAnalytics and AI with PostgreSQL - Mission Possible!
34mAnalytics and AI with PostgreSQL - Mission Possible!
50mFrom Transactions to Intelligence: Postgres for Open AI & Analyti
25mPostgres and the Artificial Intelligence Landscape
49mCan PostgreSQL Compete with Analytics Specialists?
55mDistributed In-Database Machine Leaning
47mPostgres & Beyond
48mData Horizons With Postgres