Git the NoSQL Database

on YouTube ↗39m

" summary

Claim: git can serve as a schema-less key‑value database with built‑in history, hooks, branches and easy replication; experiments include an issue tracker called gasket and libraries like grit and libgit2 (rugged bindings exist). Concrete limits: about 100 commits/sec on SSD for a single hot file, 14/sec for 1,000 flat keys, 67/sec after sharding paths; Facebook reported 4M commits 1.3M files 15GB repack took two days and simple ops ranged from 7s to 41min. Major failures: no transactions, frequent race conditions, painful merge conflicts, and poor scale for write‑heavy workloads.

AI-written summary · about

" chapters (9)

1. 0:00Introduction and motivation
2. 4:07Understanding Git plumbing
3. 7:30Data management mechanisms
4. 13:31Ruby library implementation
5. 16:16ORMs for Git
6. 17:56Benefits of Git as storage
7. 23:09Challenges and limitations
8. 28:52Scaling and infrastructure
9. 30:57Use cases and Q&A

" related talks

• 40mUsing Git as a NoSql DatabaseKenneth Truyers · DevDays Europe 2019
• 48mUsing Git as a NoSql DatabaseKenneth Truyers · Voxxed Days Bucharest 2018
• 35mGit: the stupid NOSQL databaseRick Olson · RubyConf 2010
• 22mDatMax Ogden · NodeConf ONE-SHOT Oakland 2014
• 25mManaging changes to your data with datKarissa McKelvey · EmpireNode 2015
• 52mFrom 0 $100 billion: Inside lessons on how to scale your dev workflowMichiel Michiel Toneman · Devoxx Belgium 2017
• 21mUse cases that drive the need for NoSQL databasesJeff Morris · Big Data & AI Toronto 2020
• 43mGit for Data- Data Versioning for Reproducible Data Science with DoltZach Musgrave · PyData NYC 2022