Fittl: Waiting for Postgres 18: Accelerating Disk Reads with Asynchronous I/O

Lukas Fittl writes in detailon the pganalyze blog about the asynchronous I/O capability coming with thePostgreSQL18 release.

Asynchronous I/O delivers the most noticeable gains in cloudenvironments where storage is network-attached, such as Amazon EBSvolumes. In these setups, individual disk reads often take multiplemilliseconds, introducing substantial latency compared to localSSDs.

With traditional synchronous I/O, each of these reads blocks queryexecution until the data arrives, leading to idle CPU time anddegraded throughput. By contrast, asynchronous I/O allows Postgresto issue multiple read requests in parallel and continue processingwhile waiting for results. This reduces query latency and enablesmuch more efficient use of available I/O bandwidth and CPU cycles.