Asynchronous IO in Postgres 18

Postgres 18 just dropped with a bunch of cool new toys—think UUIDv7, the Over explain module and composite index skip scans—but the real headliner is native asynchronous IO. In this episode, Hussein breaks down why async wins over traditional sync calls, revisits Postgres 17’s sync-only approach and dives into the nitty-gritty of getting non-blocking disk ops to play nice with Postgres.

He walks you through the guts of the new io_method options—worker threads, io_uring and the fallback sync mode—shares the hurdles the team overcame, and teases what’s next (hello, backend writer support and direct IO). If you’ve ever wondered how Postgres is leveling up its IO game, this is your express ticket.

Watch on YouTube