Unlock the future of PostgreSQL performance with this deep dive into Asynchronous I/O in Postgres 18—one of the most significant architectural changes in the database’s history.
This slide deck breaks down how Postgres is evolving from traditional blocking I/O to modern async processing, enabling dramatically higher throughput, better CPU utilization, and improved performance on today’s NVMe-powered infrastructure. Through clear visuals and practical examples, it explains not just what changed, but why it matters—from real-world storage constraints to the inner workings of io_uring and parallel I/O execution.
Whether you’re a database engineer, performance enthusiast, or architect designing for scale, you’ll gain actionable insights into:
- How async I/O works and why it’s a game-changer
- Key new features and configuration options in Postgres 18
- Expected performance gains and real-world impact
- What’s coming next in the Postgres I/O roadmap
Presentation Abstract
Postgres 18 adds native support for asynchronous I/O. This is the most significant change in the I/O subsystem in decades. Instead of blocking on reads, Postgres can now issue I/O requests and continue working, aligning the database engine with how modern storage actually works.
We will see what async I/O means in PostgreSQL and how it affects performance across different storage backends, such as local and remote NVMe storage options, different cloud block storage solutions, and the comparisons between synchronous and asynchronous I/O operations. Additionally, we’ll investigate changes to CPU efficiency, concurrency, and tail latency across all of them.
Lastly, we discuss the benefits and drawbacks, when to use async I/O, and how to unlock the power of Postgres in modern infrastructure.
You can download the presentation now.