zfs discuss - Mar 2010 - ZFS and Striped Mirror behavior with fixed size virtual disks

So I''m working up my SAN build, and I want to make sure it''s
going to behave the way I expect when I go to expand it.

Currently I''m running 10 - 500GB Seagate Barracuda ES.2 drives as two
drive mirrors added to my tank pool.

I''m going to be using this for virtual machine storage, and have
created fixed size disks (around 200GB per file).

If I''m reading the documentation correctly, if I add spindles to my
drive array (say 10 more drives) even though the data files on the disks
won''t change in size, the data will eventually migrate to all spindles
as it is changed and written out, thus improving performance.  Is this correct?
-- 
This message posted from opensolaris.org

On Wed, March 10, 2010 12:49, Matt wrote:
> So I''m working up my SAN build, and I want to make sure
it''s going to
> behave the way I expect when I go to expand it.
>
> Currently I''m running 10 - 500GB Seagate Barracuda ES.2 drives as
two
> drive mirrors added to my tank pool.
>
> I''m going to be using this for virtual machine storage, and have
created
> fixed size disks (around 200GB per file).
>
> If I''m reading the documentation correctly, if I add spindles to
my drive
> array (say 10 more drives) even though the data files on the disks
won''t
> change in size, the data will eventually migrate to all spindles as it is
> changed and written out, thus improving performance.  Is this correct?
I think so; though you''re using non-ZFS terminology which hence
isn''t
absolutely precise.

Let me rephrase it:  If you add more mirror vdevs to your pool named
"tank", that space will become instantly available to things that draw
on
that pool (filesystems, zvols, and I forget what else).  When new data has
to be written, there will be a preference for it going to less-full vdevs
in the pool; so as data is rewritten, usage will gradually even out across
the vdevs in the pool.  Having the data across more spindles can,
obviously, potentially increase performance, depending on what the
existing limiting factors are.

So, if that says about the same thing you said, then the answer to your
question is "yes".

-- 
David Dyer-Bennet, dd-b at dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

That is exactly what I meant.  Sorry for my newbie terminology.  I''m so
used to traditional RAID that it''s hard to shake.

That''s great to know.  Time to soldier on with the build!
-- 
This message posted from opensolaris.org

On Wed, March 10, 2010 13:32, Matt wrote:
> That is exactly what I meant.  Sorry for my newbie terminology. 
I''m so
> used to traditional RAID that it''s hard to shake.
No apology required; it''s natural that your questions will occur in the
terminology you''re familiar with.  It''s certainly hard to
shake.  I was
trying to clean up the terminology (by one of the laws of the Internet, I
presume *I* got something wrong in that post!) to verify I''d understood
correctly, and to offer the right terms back for learning.  Being rather a
pedant at  heart I''m sometimes a bit ham-tongued doing that kind of
thing,
too.
> That''s great to know.  Time to soldier on with the build!
Sounds like.  I''m very happy with my rather smaller setup.
-- 
David Dyer-Bennet, dd-b at dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info