From: Bill Davidsen <davidsen@tmr.com>
To: Raz <raziebe@gmail.com>
Cc: linux-fsdevel@vger.kernel.org,
Linux RAID Mailing List <linux-raid@vger.kernel.org>,
linux-xfs@oss.sgi.com, linux-aio@kvack.org,
"linux-ide@vger.kernel.org" <linux-ide@vger.kernel.org>
Subject: Re: How to configure 36 disks ?
Date: Mon, 23 Mar 2009 11:35:08 -0400 [thread overview]
Message-ID: <49C7AC2C.9090307@tmr.com> (raw)
In-Reply-To: <5d96567b0903230659t734677a3pb4fd77cccb54008b@mail.gmail.com>
Raz wrote:
> Hello
> I need to configure 3xDAS'es, each with 12 disks.
> All three DAS'es are connected to a single machine.
> I have the following requirements (in this order of importance)
> from the storage:
>
> 1. redundancy.
> having two disks failing in one raid5 breaks the entire raid. when
> you have 30TB storage
> it is a disaster.
>
> 2. performance.
> My code eliminates Linux raid5/6 write penalty. I managed to do by
> manipulating xfs and patching linux raid5 a bit.
>
> 3. modularity ( a "grow" and it will be nice to have "shrink" )
> file system and volume must be able to grow. shrinking is possible
> by unifying multiple file systems
> under unionfs or aufs.
>
> 4. Utilize storage size.
>
> I assume each disk is 1TB.
>
>
___ snip ___
> Any other ideas ?
Yes, you have the whole solution rotated 90 degrees. Consider your
original solution #2 below... You have no redundancy if one whole DAS
box fails, which is certainly a possible failure mode. If you put the
RAID0 horizontally, two arrays size six in each DAS, then RAID6
vertically, if one DAS fails completely you still have a functioning
system, and the failure results for individual drives remains about the
same, while the rebuild time will be longer.
Solution #2
raid0
DAS1: raid6: D,D,D,D,D,D |
raid6: D,D,D,D,D,D |
|
DAS2: raid6: D,D,D,D,D,D | xfs.
raid6: D,D,D,D,D,D |
|
DAS3: raid6: D,D,D,D,D,D |
raid6: D,D,D,D,D,D |
In addition, you can expand this configuration by adding more DAS units.
This addresses several of your goals.
In practice, just to get faster rebuild as the array gets larger, I
suspect you would find it was worth making the horizontal arrays RAID5
instead of RAID0, just to minimize time to full performance.
--
bill davidsen <davidsen@tmr.com>
CTO TMR Associates, Inc
"You are disgraced professional losers. And by the way, give us our money back."
- Representative Earl Pomeroy, Democrat of North Dakota
on the A.I.G. executives who were paid bonuses after a federal bailout.
next prev parent reply other threads:[~2009-03-23 15:35 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-03-23 13:59 How to configure 36 disks ? Raz
2009-03-23 14:12 ` Emmanuel Florac
2009-03-23 15:35 ` Bill Davidsen [this message]
2009-03-23 16:02 ` Jon Hardcastle
2009-03-23 16:22 ` Mark Lord
2009-03-23 16:23 ` Christopher Smith
2009-03-23 16:28 ` Raz
2009-03-23 16:45 ` Greg Freemyer
2009-03-23 18:32 ` Mikael Abrahamsson
2009-03-24 19:38 ` Goswin von Brederlow
2009-03-25 12:14 ` Drew
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=49C7AC2C.9090307@tmr.com \
--to=davidsen@tmr.com \
--cc=linux-aio@kvack.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-ide@vger.kernel.org \
--cc=linux-raid@vger.kernel.org \
--cc=linux-xfs@oss.sgi.com \
--cc=raziebe@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).