Achieving 11M IOPS and 66 GB/S IO on a Single ThreadRipper Workstation

Re:Achieving 11M IOPS and 66 GB/S IO on a Single ThreadRippe... by 1996 on Saturday January 30, 2021 @11:07AM (#25970147)

> for most flash-based SSDs, there's no reason to set ashift to anything less than 12 (corresponding to 4kB blocks).

matching the page size?

> the underlying media page size is usually on the order of 16kB

I'd say that's a good reason to set ashift=14 as 2^14=16kb

[ Parent | on HN ]

Re:Achieving 11M IOPS and 66 GB/S IO on a Single ThreadRippe... by wtallis on Saturday January 30, 2021 @02:09PM (#25971949)

There are downsides to forcing the OS/FS to always use larger block sizes for IO. You might simply be moving some write amplification out of the SSD and into the filesystem, while losing some performance in the process. Which is why it really depends on your workload, and to some extent on the specific SSD in question. I'm not convinced that ashift=14 is a sensible one size fits all recommendation, even if we're talking only about recent-model consumer-grade NAND SSDs.

[ Parent | on HN ]