| From: | Michael Paquier <michael(at)paquier(dot)xyz> |
|---|---|
| To: | Andres Freund <andres(at)anarazel(dot)de> |
| Cc: | Robert Haas <robertmhaas(at)gmail(dot)com>, Peter Geoghegan <pg(at)bowt(dot)ie>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: adding 'zstd' as a compression algorithm |
| Date: | 2022-02-17 01:48:13 |
| Message-ID: | Yg2pXfmguCKTK935@paquier.xyz |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Tue, Feb 15, 2022 at 06:24:13PM -0800, Andres Freund wrote:
> For backups it's pretty obviously zstd imo. At the lower levels it achieves
> considerably higher compression ratios while still being vastly faster than
> gzip. Without even using the threaded compression support the library has.
Noted.
> For something like wal_compression it'd be a harder question.
FWIW, I have done some measurements for wal_compression with zstd, as
of:
https://www.postgresql.org/message-id/YMmlvyVyAFlxZ+/H(at)paquier(dot)xyz
The result is not surprising, a bit more CPU for zstd with more
compression compared to LZ4, both outclassing easily zlib. I am not
sure which one would be more adapted as default as FPI patterns depend
on the workload, for one, and this is just one corner case.
--
Michael
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Andres Freund | 2022-02-17 01:50:52 | Re: O(n) tasks cause lengthy startups and checkpoints |
| Previous Message | Jeevan Ladhe | 2022-02-17 01:46:22 | Re: refactoring basebackup.c |