Do You Even Need a Database?
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A practical benchmark comparing flat-file storage strategies (linear scan, in-memory hash map, binary search on disk) against SQLite across Go, Bun, and Rust HTTP servers. Results show in-memory maps reach ~97-169k req/s, binary search on disk ~40k req/s, and SQLite ~25k req/s — all consistent regardless of dataset size (10k to 1M records). The post argues most applications don't need a database early on, since even SQLite on a single server can handle ~90 million DAU. It also outlines when you genuinely need a database: dataset exceeds RAM, multi-field queries, joins, multi-process writes, or ACID transactions.
Table of contents
The setupApproach 1: Read the file every timeApproach 2: Load it into memoryApproach 3: Binary search on diskThe benchmarkThe resultsWhat does 25,000 requests per second actually mean?When do you actually need a database?Sort: