Date: 2025-05-15
Time: 10:30–11:20
Room: Breakout Room 1
Level: Intermediate
I've been, on and off, working on adding AIO support to Postgres for ~5 years. While, hopefully, core pieces will land in PG 18, this is my longest development project so far.
It seems worth trying to analyze what lead to this long development timeframe and to see if any lessons can be learned. AIO is a large architectural change and there are many other large architectural changes we need to make for postgres to somewhat keep up with the needs of our users.
I'll try to analyze what went wrong from a few angles, including:
1) technical / architectural 2) procedural (from both a project and personal angle) 3) personal / motivational
To make sure this won't be too depressing, I'll also include some things that went ok, too :)