Use SQLite in production
summary
Argues that SQLite is a viable production DB: measured client reports show 800–8,963 SQLite files per laptop (roughly 2,000 average) and roughly 806,000 DBs at the conference, simple queries run in microseconds, a Wagtail test hits ~50 page updates/sec on one Unicorn worker, and a demo shows 85ms vs 150ms vs Postgres. Recommends enabling journal_mode=WAL and tuning synchronous, using lightstream to stream WAL to S3 for point‑in‑time and global replication, and using sqlite‑vss or dataset for vector search.
AI-written summary · about
chapters (11)
- 0:00The case for SQLite
- 2:05Ubiquity and reliability
- 5:57Simplicity and philosophy
- 8:31Vector search capabilities
- 10:53Serverless nature
- 14:04Performance and latency
- 17:30Addressing concurrency
- 20:03Data integrity and typing
- 20:52Recovery and reliability
- 22:33Scaling and final advice
- 24:57Q&A session
related talks
01h 07mHow (and why) to run SQLite in production
45mSQLite on Rails: Everything you need to know
31mSQLite, the World's Most Ubiquitous Yet Little-Known Database
01h 56mSQLite on Rails: From rails new to 50k concurrent...
39mBuilding Production Applications Using Go
19mSQLite - The most overlooked database
01h 56mSQLite on Rails: From rails new to 50k concurrent users and everything in between
23mFANTASTIC DATABASES AND WHERE TO FIND THEM