Strange. SQLite runs most of the world but I haven't seen it mentioned in this thread once.

90%+ of software products could easily get by with SQLite as a data persistence mechanism if the developers can be brought to a state of humility regarding the scale of their projects.

I'm seeing comments in here talking about 200 vs 20 milliseconds. You know what's even faster? Putting the DB engine in-proc and dropping the network from the equation altogether. With NVMe storage, you can reliably complete SQLite transactions in tens to hundreds of microseconds with the appropriate journaling mode enabled. You'll never get this kind of latency bound from something that lives in another process or computer.

Another big open source DB I'd like to hear about more often is Firebird (https://firebirdsql.org/) which was forked from the Interbase code released by Borland (don't remember how they were called at the time). Four choices of connection model (process-by-connection, thread-by-connection, some weird mix, and in-process), full ANSI SQL, runs on all major platforms, and uses single file databases. Seems to be the best of SQLite (inmem, single file), MySQL (lots of connections fast) and Postgres (standard, robust) to me, in a single package. But for whatever reason it seems to be completely off the radar on HN...