Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 27 Nov 2013 14:17:40 +0200
From:      Daniel Kalchev <daniel@digsys.bg>
To:        freebsd-fs@freebsd.org
Subject:   Re: Performance difference between UFS and ZFS with NFS
Message-ID:  <5295E2E4.8050506@digsys.bg>
In-Reply-To: <CAM=5oeDR3rSmxpffJnTFSkXMxRY0a0gz-NnogoQLhc4ih06gPw@mail.gmail.com>
References:  <2103733116.16923158.1384866769683.JavaMail.root@uoguelph.ca> <9F76D61C-EFEB-44B3-9717-D0795789832D@gmail.com> <CAM=5oeAF2gfccrGNdbApUDpqRae4OQjZ7oaZZi4y1j%2BsF6PsTw@mail.gmail.com> <5969250F-0987-4304-BB95-52C7BAE8D84D@gmail.com> <CAM=5oeBmCAq9unFGC2CBoJ3rZMm9MtDw1DWkFpo2ZqQtx3G%2B=Q@mail.gmail.com> <18391B9C-2FC4-427B-A4B6-1739B3C17498@gmail.com> <CAM=5oeAXiRn2aHvNPuZRPFJp6G45OqdQEDsz2_xGobCUHJp_VQ@mail.gmail.com> <CAM=5oeAybYZsn-TQTvfYO3Kx-chZULeYUEhg9fRMi-CFX0aYBA@mail.gmail.com> <B7C29F362A5045B48BF4CE223E7CE0AD@multiplay.co.uk> <CALfReyfDWTsgXf8JX-eNYOTFdkcYovUf0q=zGQPDUsEUi4nqbw@mail.gmail.com> <CAM=5oeAPG0Vfa-fT1t=Vc8FS_jJGBiR=yq0wuUmEZYLYB%2Br7RA@mail.gmail.com> <DD94A3B19E23453189D4A0C5AB58618A@multiplay.co.uk> <F1C5007B61404180A44CA5719C859EDE@multiplay.co.uk> <CAM=5oeDR3rSmxpffJnTFSkXMxRY0a0gz-NnogoQLhc4ih06gPw@mail.gmail.com>

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

On 25.11.13 21:24, Eric Browning wrote:
> Steven,
>
> I've tried to 4K align these SSDs with gnop but they are currently ashift
> 9.  Pool layout is just 4 drives in a zfs stripe.  I've also tried raidz1
> with no noticeable performance impacts other than a loss of space for
> parity.
>

ashift=9 with most (all?) SSDs is a big no-no! You really should make 
that pool ashift=12 (at least) and have it 4k aligned (partition).

Especially for writes, an properly aligned 4 drive SSD stripe should be 
way faster.

Here is what I get from ashift=12 raidz1 4 SSD drive pool

# dd if=/dev/zero of=zero bs=1m count=1k
1024+0 records in
1024+0 records out
1073741824 bytes transferred in 1.817027 secs (590933382 bytes/sec)

Unfortunately, your only option is dump / recreate pool / restore. (zfs 
send/receive is an option too -- especially if you have another set of 
drives)


Daniel



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5295E2E4.8050506>