TL;DR

Postgres 18 just dropped a bunch of cool features—think native UUIDv7, an Over explain module and composite index skip-scans—and the headline act is true asynchronous I/O. In this Backend Engineering Show episode, Hussein walks you through why sync I/O can be a blocker, how async I/O works under the hood, and why Postgres 18’s new worker thread and Linux io_uring modes are game-changers compared to the old sync approach.

He also zeroes in on what’s next: better support for backend writers, direct I/O, and fine-tuning the worker mode. If you’ve been itching for nonblocking disk operations in Postgres, this deep dive is your one-stop guide to why async I/O finally arrived (and why it still has room to grow).

Watch on YouTube