re: ASM performance on storage RAID 1 or RAID 10 devices

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Guys,
	I'm wondering if any of you have had a chance to test ASM
performance
	against a RAID 10 vs RAID 1 device.  This is the concern:

	ASM does 1 MB file extent stripping, and so does it make
sense to use a raid
	device that is also stripped .. specially if the storage
cannot strip with
	such large strip size?

	The question would be, does one get better IOPS and MB/s
with ASM against
	a raid device that is just mirrored or a raid device that
is mirrored and stripped with
	a stripe size that is a multiple of 1MB?

	I spoke to an Oracle architect way back about this, and he
said it doesn't
	really matter, but that if you can strip at the storage
level, then you should!

	Does anyone have experience testing this? The IO workload
that I'm concerned with
	is lots of small random ios + large read-only sequential
ios ?
	I'm familiar with Oracle's tool Orion for testing IO, but
I'm wondering if anyone has tested this out ?

	Any feedback is greatly appreciated!
	BTW, my environment is ( SuSe Enterprise 10 x86_64 +
10.2.0.3 + storage (DS4400 Disk Systems (FAStT 700))

- -peter
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


iD8DBQFGFPwIoyy5QBCjoT0RAvxZAJ9uvdbaHjlOj3SxIdatPNi2jcVeCQCf
Z2Zi
jocFhwvkvNIWNKAM/TCfNN0=
=CtIA
-----END PGP SIGNATURE-----

-- 
To unsubscribe, email: suse-oracle-unsubscribesuse.com
For additional commands, email: suse-oracle-helpsuse.com
Please see http://www.suse.com/oracl
e/ before posting

Hi Peter,

I am running such a setup on SLES9 64bit 10.2.0.x on a
DS4800.

Striping over striped devices scattered performance very
much. My best 
solution was to do every raid operation via ASM. Now I use
arrays like disks: 
RAID5 arrays, stripe over them by ASM (more arrays in one
failgroup) and 
mirror them to another bunch of arrays (normal redundancy).
If one fails, I 
replace it completely within ASM (not physically). 

Amazon does so with cheap, dumb raid subsystems, and they
are running very 
fast on it.

I do my performance monitoring by storage delivered tools,
and grid 
control/database control.

Don't know if it helps. You are welcome to write back, since
I am interested 
in different experiences all the time. These systems are
seldom found 

Am Donnerstag 05 April 2007 15:39 schrieb Peter Santos:
> Guys,
> 	I'm wondering if any of you have had a chance to test
ASM performance
> 	against a RAID 10 vs RAID 1 device.  This is the
concern:
>
> 	ASM does 1 MB file extent stripping, and so does it
make sense to use a
> raid device that is also stripped .. specially if the
storage cannot strip
> with such large strip size?
>
> 	The question would be, does one get better IOPS and
MB/s with ASM against
> 	a raid device that is just mirrored or a raid device
that is mirrored and
> stripped with a stripe size that is a multiple of 1MB?
>
> 	I spoke to an Oracle architect way back about this,
and he said it doesn't
> 	really matter, but that if you can strip at the
storage level, then you
> should!
>
> 	Does anyone have experience testing this? The IO
workload that I'm
> concerned with is lots of small random ios + large
read-only sequential ios
> ?
> 	I'm familiar with Oracle's tool Orion for testing IO,
but I'm wondering if
> anyone has tested this out ?
>
> 	Any feedback is greatly appreciated!
> 	BTW, my environment is ( SuSe Enterprise 10 x86_64 +
10.2.0.3 + storage
> (DS4400 Disk Systems (FAStT 700))
>
> -peter

-- 
Freundliche Grüße

i.A.
Martin Klier
Systemadministration/Datenbanken
------------------------------------------------------------
------
A.T.U Auto-Teile-Unger
Handels GmbH & Co. KG