Breakout Session
summary
ToroDB runs on top of PostgreSQL and claims a MongoDB-compatible API while storing data relationally rather than in jsonb, splitting documents into per-level tables and a structure table that describes shapes so thousands of structures become separate tables. It uses PostgreSQL logical decoding to act as a MongoDB slave, is AGPL and Java-based, and supports transactions and clean reads. Benchmarks on a 60 GB Amazon dataset show much lower disk use and faster SQL analytics; one MongoDB query fails to complete.
AI-written summary · about
chapters (26)
- 0:00Start
- 2:00ToroDB in one slide
- 2:31How to make a NoSQL database
- 3:54Toro DB's Document 2 Relational transformation
- 6:32Toro DB storage internals
- 10:52Why not isonb?
- 12:29Metadata repetition
- 13:06ToroDB storage and I/O savings
- 13:54ToroDB: query "by structure"
- 16:56Data normalization
- 17:38MongoDB WP and API
- 18:53ToroDB: native SQL
- 20:57Toro DB VIEWS
- 22:34VIEWS, Toro DB from any SQL tool
- 23:44Replication protocol choice
- 25:09MongoDB's replication protocol
- 25:50MongoDB replication implementation
- 27:11MongoDB's oplog & tailable cursors
- 32:22Benchmark
- 32:51Disk usage
- 33:34Queries: which one is easier?
- 34:44Query times
- 36:13Atomic operations
- 37:22Cheap single-node durability
- 38:39"Clean" reads
- 40:15Replication dirty and stale reads ("call me maybe")
related talks
01h 04mToroDB: NoSQL & relational Java database with SQL analytics
48mToroDB scaling PostgreSQL like MongoDB
48mToroDB scaling PostgreSQL like MongoDB
33mToroDB Stampede: open-source BI connector for MongoDB
53mPostgres vs Mongo
48mThe NoSQL Way in PostgreSQL
53mNoSQL Best Practices for PostgreSQL
14mJSON data structure – the hidden gem inside your database engine