Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 3 Sep 2024 09:33:32 +0300
From:      Anthony Pankov <anthony.pankov@yahoo.com>
To:        Daniel Ebdrup Jensen <debdrup@freebsd.org>, freebsd-hackers@freebsd.org
Subject:   Re: makefs -t ffs makes too large image
Message-ID:  <1876116975.20240903093332@yahoo.com>
In-Reply-To: <afdudyoicqjkpwlpvyukcruvgaf5amrlnmnczi7ahlmtjng2em@swopo3io4pgz>
References:  <1781435895.20240822135231.ref@yahoo.com> <1781435895.20240822135231@yahoo.com>  <d3eb46f1-29cc-4fa0-a346-86e1c5434834@quip.cz> <334195982.20240822182405@yahoo.com> <70579701-7457-46b1-9329-a51eddec51de@aetern.org> <afdudyoicqjkpwlpvyukcruvgaf5amrlnmnczi7ahlmtjng2em@swopo3io4pgz>

next in thread | previous in thread | raw e-mail | index | archive | help
Hello Daniel,

My impression that compression makes ZFS slower on VM. May be this is the only case of not so-well-optimized VM such as QEMU on Windows.

Sunday, September 1, 2024, 11:50:16 PM, you wrote:

> On Fri, Aug 23, 2024 at 12:39:26AM UTC, Yuri Pankov wrote:
>>Anthony Pankov wrote:
>>> Hello Miroslav,
>>>
>>> You are  genius!
>>>
>>> But the situation is a very frustrating. It is a default system and I did nothing  to turn the compression on.
>>> So I was absolutely sure that compression is off.
>>>
>>> I'm sorry.
>>> Nevertheless having compression on by default is a very weird decision and is fully unexpected for me. I've never seen a big warning about default value of this vital parameter will be inverted.
>>>
>>> On 12 -STABLE:
>>>
>>>  # zfs get compression
>>> NAME                       PROPERTY     VALUE     SOURCE
>>> ps2                        compression  off       default
>>>
>>> On 14-STABLE
>>>
>>>  # zfs get compression
>>> NAME                          PROPERTY     VALUE           SOURCE
>>> tank                          compression  on              default
>>> tank/bsdsrc                   compression  on              default
>>
>>It came in with the following openzfs commit and probably no one really
>>noticed as installer turns on compression by default, so I was going to
>>say it was always that way until I looked up the change :-)
>>
>>commit 56fa4aa96eb3875f254e93eaef646ea20ba187f9
>>Author: Rich Ercolani <rincebrain@gmail.com>
>>Date:   Thu Mar 3 13:43:38 2022 -0500
>>
>>    Default to ON for compression
>>
>>    A simple change, but so many tests break with it,
>>    and those are the majority of this.
>>
>>    Reviewed-by: George Melikov <mail@gmelikov.ru>
>>    Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
>>    Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
>>    Closes #13078
>>
>>And it looks like it's in FreeBSD starting with 14.0:
>>
>>$ git branch -a --contains 56fa4aa96eb3875f254e93eaef646ea20ba187f9
>>* main
>>  remotes/origin/HEAD -> origin/main
>>  remotes/origin/main
>>  remotes/origin/pull/956/merge
>>  remotes/origin/releng/14.0
>>  remotes/origin/releng/14.1
>>  remotes/origin/stable/14
>>

> Hi folks,

>       It's maybe also worth mentioning that compression by default makes
>       ZFS faster.

>       Allan Jude had some numbers for the default (lz4) in the
>       presentation he did while implementing zstd[1].

>       It can be a bit surprising though, but hopefully it's a good one.;)

> Yours,
> Daniel Ebdrup Jensen

> 1: https://papers.freebsd.org/2018/bsdcan/jude-zfs_zstd/


-- 
Best regards,
Anthony




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1876116975.20240903093332>