Re: Is there a performance advantage of adding a second parity set?

Back a few years ago, I ran the statistical calculations comparing 1
RAID set vs. two...
http://groups.google.com/group/comp.sys.ibm.as400.misc/browse_thread/thread/fdbaa180e9f62b33/f3a118a03f3a1040

The end result, you're much safer with 2 than you are with one, but
after 2 there's not as big an improvement in safety.

Also note there's some performance numbers thrown about in the above
thread, 5-10% cost of 2 RAID vs. 1...but given the number of years ago
and the disk hardware in question, I don't know that I'd consider that
a current cost.

HTH,

Charles


On Mon, Feb 15, 2010 at 10:26 AM, John Jones <chianime@xxxxxxxxx> wrote:

True, that can happen and it happened once to me in the past.  It was an
ugly restore.

But if you lost two drives at once and they were in different parity sets
you would not have gone down.  Multiple parity sets makes you more
reliable.  If you went with groups of 8 drives in a parity set and had 3
sets/24 drives, you could lose 1 drive in each set - 3 total - and still be
running.  Reduce the parity sets to 4 drives per and you have 6 sets and
could again lose 1 drive per set and stay running only this time it's 6
drives that you could lose without going down.  Losing 2 drives in the same
set will still cause a crash, however.  Taken to the extreme is mirroring
where every drive has a "parity" match and you could potentially lose up to
half your drives and still be running.  It has to be the right half, though
..

As you can guess, even in a mirrored environment you can go down if you lose
just 2 drives.  That would happen if the 2 drives are mirrors of each other.

Each added parity sets makes you incrementally able to withstand more drive
failures.  It isn't perfect since the added failures can still be in the
same set but the odds do get better.

On Mon, Feb 15, 2010 at 9:14 AM, Jeffry A. Kennedy
<jkennedy@xxxxxxxxxxxxx>wrote:

We have 3 parity sets on our 520 (HA Box) an we just lost two 35 Gig Drives
in the same parity set..

-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx [mailto:
midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of John Jones
Sent: Monday, February 15, 2010 8:42 AM
To: Midrange Systems Technical Discussion
Subject: Re: Is there a performance advantage of adding a second parity
set?

If you add 6 drives at once the parity should spread across four of them as
you presume.

I don't know that the performance difference would be measurable but I
would hazard a guess that 2 parity sets would be better than one.  With just
one set those 8 parity drives will be spending more and more time performing
parity writes for the additional drives; on average they will be busier
which could induce delays when the OS wants to read from them.  More parity
sets spreads the parity writes across more drives sort of like how the i
scatter-loads data to begin with.  The number of parity writes is unchanged
but the busy-ness of the drives that have parity duty is reduced.

Also, a second parity set increases your reliability as you could lose one
drive in each set and still be operational.

On modern RAID cards, what is the maximum number of drives per parity set
anyway?  I thought it topped out at 15 or 18.

On Mon, Feb 15, 2010 at 8:18 AM, <rob@xxxxxxxxx> wrote:

We have a system with 17 drives.  We are going to add six 70GB drives.
If I add them two at a time they will add them to the existing parity
set. If I add them all at once it will start a new parity set.
Starting a new parity set effectively ties up one of the 70GB drives
to raid striping. We really don't need the space though - we just have
70GB drives laying around.  Without the new drives we are only at 50%
used.  The existing parity set has the striping across 8 drives.  I
take it the new set would have the striping across 4 drives?

These six 4327's would go into a 5787.  Into P1-D1 through P1-D6.
P1-D7 through P1-D12 are 4328's.  P2-D1 through P2-D5 and P2-D7
through
P2-D11 are 4328's.  Leaving P2-D6 and P2-D12 open.

Rob Berendt
--
Group Dekko Services, LLC
Dept 01.073
Dock 108
6928N 400E
Kendallville, IN 46755
http://www.dekko.com

--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing
list To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe,
unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take
a moment to review the archives at
http://archive.midrange.com/midrange-l.

--
JJ
4 Out of 3 people have trouble with fractions.
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe,
unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a
moment to review the archives at http://archive.midrange.com/midrange-l.

--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.

--
JJ
4 Out of 3 people have trouble with fractions.
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.