‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Saturday, April 24th, 2021 at 21:03, Rory Campbell-Lange <rory@xxxxxxxxxxxxxxxxxx> wrote: > On 24/04/21, Christophe Pettus (xof@xxxxxxxxxxxx) wrote: > > > > On Apr 24, 2021, at 11:27, Simon Connah simon.n.connah@xxxxxxxxxxxxxx wrote: > > > > > > I'm curious, really. I use btrfs as my filesystem on my home systems and am setting up a server as I near releasing my project. I planned to use btrfs on the server, but it got me thinking about PostgreSQL 13. Does anyone know if it would have a major performance impact? > > > > This is a few years old, but Tomas Vondra did a presentation comparing major Linux file systems for PostgreSQL: > > > > https://www.slideshare.net/fuzzycz/postgresql-on-ext4-xfs-btrfs-and-zfs > > I guess btrfs and zfs on linux performance might have improved somewhat since Tomas' analysis in 2015. > > Personally I've been a keen personal user of btrfs for over 5 years for its snapshot support, transparent compression and bitrot detection. However I can't think of a reason to use it for a production server. It's slower than ext4 and xfs, postgresql's dumping and streaming are probably better bets than snapshots for backup, and relatively few others are likely to be comfortable administering it. But maybe I'm missing something. > > Phoronix run btrfs benchmarks from time-to-time. See https://www.phoronix.com/scan.php?page=search&q=Btrfs > > Rory I should have been more explicit in my original post. I use OpenSUSE Tumbleweed and btrfs is the default filesystem. I also believe SUSE Linux Enterprise Server uses btrfs as their default which seems to indicate that at least one enterprise-grade Linux distribution trusts btrfs for their default install. I'd love to do some performance testing myself but the only way I could do it would be inside virtual machines which would make the numbers meaningless. I'll certainly do some reading on the subject and look at that presentation that was mentioned.
Attachment:
signature.asc
Description: OpenPGP digital signature
