Interesting !
the debian blog saying that
"
but only users who are interested in debugging future bugs should use transparent compression at this time"
makes me feel its less urgent to test this option...
On Mon, Apr 26, 2021 at 3:43 PM Rory Campbell-Lange <rory@xxxxxxxxxxxxxxxxxx> wrote:
On 26/04/21, Marc Millas (marc.millas@xxxxxxxxxx) wrote:
> compression ?
>
> I am currently working on a project to move an oracle db to postgres.
> The db is 15 TB.
> with Oracle compression it does use 5 TB of disk space.
>
> If we cannot compress the whole thing, the project loses its economic base.
> (added 10 TB for prod, 10TB for pre-prod, 10TB for testing dev, ...)
>
> we do test zfs, and we will give a try to btrfs.
I've been using btrfs with lzo compression for several years on my
personal laptop and some non-critical backup systems with no trouble.
(In fact btrfs has helped us recover from some disk failures really
well.) While I run postgresql on my machine it is for light testing
purposes so I wouldn't want to comment on its suitability for
production.
There are some differences reported here between lzo and zlib
compression performance for Postgresql:
https://sudonull.com/post/96976-PostgreSQL-and-btrfs-elephant-on-an-oil-diet
zstd compression support for btrfs is reported on by Phoronix here:
https://www.phoronix.com/scan.php?page=article&item=btrfs-zstd-compress&num=2
The compression page of the btrfs wiki is here:
https://btrfs.wiki.kernel.org/index.php/Compression
You might want to armor yourself for possible problems by reading the
Debian btrfs wiki page: https://wiki.debian.org/Btrfs
If you test your workload please let us know your results.
Rory
