We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Ilya Kosmodemiansky: Linux IO internals for PostgreSQL administrators in 2023 (PGConf.EU 2023)
Get a deep dive into Linux IO internals and learn how to optimize input-output performance for PostgreSQL administrators, covering NVMe, bulk MQ scheduling, buffered I/O, and more.
- Linux IO internals are complex and have evolved significantly over time.
- Database administrators should be aware of the importance of optimizing input-output (IO) for their workloads.
- NVMe and bulk MQ scheduling are key concepts that can improve IO performance.
- Postgres uses buffered I/O by default, but can benefit from direct I/O in some cases.
- SSDs are highly parallel-capable and can provide better performance than traditional disk drives.
- Linux kernels have developed to be NVMe-aware and can provide better support for SSDs.
-
Input-output monitoring is essential for performance optimization and Postgres provides extensions like
pgstat.iofor this purpose. - Writing-ahead log and buffer management are crucial for maintaining data consistency and safety.
- Linux IO stack needs to be optimized for high-performance workloads, especially for databases.
- Practitioners should understand the difference between throughput and latency and how to optimize for both.
- The talk provides a high-level overview of Linux IO internals and suggests resources for further learning.