[geeks] Impressive...

velociraptor velociraptor at gmail.com
Fri Mar 6 17:04:50 CST 2009


On Wed, Mar 4, 2009 at 7:01 PM, Francois Dion <francois.dion at gmail.com>
wrote:
> On Wed, Mar 4, 2009 at 8:43 PM, Nadine Miller <velociraptor at gmail.com>
wrote:
>> On Mar 4, 2009, at 5:22 PM, Francois Dion wrote:
>>>
>>> How are you thinking of using the 2.5"? I use 2x 3.5 for the os
>>> (bootable zfs) and /opt and other anciliaries, and two mirrors, for 6
>>> drives total. I used to have oracle on a clariion but zfs and sata is
>>> faster. If I needed more write performance, I'd add a pair of SSD for
>>> the ZIL. But even with multiple ATSC HDTV streams recording and
>>> playing at the same time I still have plenty of wiggle room without
>>> that, so...
>>
>> I'd like to hear more about that Oracle Clariion issue if you are willing
> to
>> share, Francois?
>
> This is at home... yeah, I know, I should seek treatment :)
>
> It was an SGI branded Clariion, with Thor controllers and it was
> connected to my E5500. When I found a v480 fully loaded for less than

Do you know how that compares to current models? Clariion 380, maybe?
Just wondering.

>> Why not RAID5?  Is the performance that dramatically different?  If that's
>> the case, I'm going to have to re-think my hardware plans.
>
> Well, he was saying he hates buying 4 disks at once. 2 is kind of a
> minimum, so mirror. If I had 8 disks, that would be a different
> matter, cost differential is much greater, I'd go for raidz2. Put 4
> disks on each controllers. 3+1, 3+1 for 6+2 parity.
>
> At 4 disks, it's a toss between capacity and performance, you'd want
> to experiment with your load. I do mostly reads, and I get a better
> throughput (you could test with bonnie++). But raidz (i'm assuming you
> are not talking raid5 as 1 lun that zfs sees as 1 disk) has some
> limitations that I find tips the balance to 2 mirrors instead of a 3+1
> raidz, which is what I did.
>
> You have to realize that short of plain mirrors, you cant really grow
> properly a raidz or raidz2. You cant reduce the size either. That's
> why I suggested adding pools in mirrored pairs, until ZFS can properly
> grow/shrink/restripe transparently. Currently you have to backup to
> tape recreate your new pool and reload from tape (or other backup).
> You can scrub / resilver live so I have good hopes that this feature
> will eventually make its way in solaris 10 and be usable live.

I need to educate myself more, obviously.  I didn't realize you could
not grow the pools transparently.  My original plan was to take two
new disks and stripe them, move the data to those, then use the
existing 3 disks to start the zpool, transfer the data to zfs, then
put the two new disks into the zpool and grow it.  This will have to
be revised, and I'm going to need to reconsider the disks I use for
the project.

What's the latest chatter on the Seagate firmware debacle?  Has it settled
out?

=Nadine=



More information about the geeks mailing list