zfs discuss - Sep 2010 - ZFS on solid state as disk rather than L2ARC...

Has anyone done much testing of just using the solid state devices  
(F20 or F5100) as
devices for ZFS pools?  Are there any concerns with running in this  
mode versus using
solid state devices for L2ARC cache?

Second, has anyone done this sort of testing with MLC based solid  
state drives?
What has your experience been?

Thanks in advance!

Brad

Brad Diggs | Principal Security Sales Consultant
Oracle North America Technology Organization
16000 Dallas Parkway, Dallas, TX 75248
Tech Blog: http://TheZoneManager.com
LinkedIn: http://www.linkedin.com/in/braddiggs

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100915/6ac9a056/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PastedGraphic-2.tiff
Type: image/tiff
Size: 9062 bytes
Desc: not available
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100915/6ac9a056/attachment-0001.tiff>

We actually did some pretty serious testing with SATA SLCs from Sun directly
hosting zpools (not as L2ARC).  We saw some really bad performance - as though
there were something wrong, but couldn''t find it.

If you search my name on this list you''ll find the description of the
problem.

--m

m a t t    c o w g e r
sr. systems engineer | internal services / techops
salesforce.com
+1 206 262 8078 (office)
+1 415 401 5930 (mobile)
mattkcowger (aim  |  yahoo)

From: zfs-discuss-bounces at opensolaris.org [mailto:zfs-discuss-bounces at
opensolaris.org] On Behalf Of Brad Diggs
Sent: Wednesday, September 15, 2010 12:27 PM
To: zfs-discuss at opensolaris.org
Subject: [zfs-discuss] ZFS on solid state as disk rather than L2ARC...

Has anyone done much testing of just using the solid state devices (F20 or
F5100) as
devices for ZFS pools?  Are there any concerns with running in this mode versus
using
solid state devices for L2ARC cache?

Second, has anyone done this sort of testing with MLC based solid state drives?
What has your experience been?

Thanks in advance!

Brad
[cid:image001.png at 01CB54D6.1F2EB3C0]
Brad Diggs | Principal Security Sales Consultant
Oracle North America Technology Organization
16000 Dallas Parkway, Dallas, TX 75248
Tech Blog: http://TheZoneManager.com<http://TheZoneManager.com/>
LinkedIn: http://www.linkedin.com/in/braddiggs

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100915/c653686a/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 4825 bytes
Desc: image001.png
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100915/c653686a/attachment.png>

On Wed, Sep 15, 2010 at 12:27 PM, Brad Diggs <Brad.Diggs at oracle.com>
wrote:
> Has anyone done much testing of just using the solid state devices (F20 or
> F5100) as
> devices for ZFS pools?  Are there any concerns with running in this mode
> versus using
> solid state devices for L2ARC cache?
>
Depending on the configuration (raidz vs. mirror) you can wind up issuing
unaligned writes. Current SSDs work best when writes are aligned on a 4k
boundary, which raidz won''t always do.

-B

-- 
Brandon High : bhigh at freaks.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100915/e2fd6ebb/attachment.html>

14x 256gb MLC SSDs in a raidz2 array have worked fine for us. Performace seems
to be mostly limited by the raid controller in operating in JBOD mode.

Raidz2 allows sufficient redundancy to replace any MLC drives that develop
issues and when you have that many consumer level SSDs, some will develop issues
until you get all the weak ones weeded out.

If I had more space in the box or didn''t need quite as much space I
would have split the array into 2 or more striped raidz2 arrays.

Your mileage may vary.
-- 
This message posted from opensolaris.org