Re: RAC and ASM disk layout

  • From: "goran bogdanovic" <goran00@xxxxxxxxx>
  • To: HELMUT.DAIMINGER@xxxxxx
  • Date: Sun, 11 Jun 2006 12:37:13 +0200

Hi Helmut,

"all LUNs will end up on the same physical devices anyway. So it might
be harmful having ASM to try and balance I/O across several LUNs which
all reside on the same physical devices"
This is true, and if you do that you will get a lot of disk contention.
As of RAID-5, the general rule of thumb is to deploy RAID-5 where cost
of storage is critical, performance is not a primary goal, and
applications with primary read operations like DWH.
Since ASM is based on the SAME technology and is LVM itself, using LVM
with ASM is not a good idea since it would complete a duplicating of
functionality, overhead and complexity.
ASM striping with a hardware RAID-0 is OK if configured correctly. In
that case use 1Mb stripe size. The smaller than 1Mb stripe size is
inefficient => breaks down the 1Mb I/O into multiple smaller I/Os and
because a physical disk constraints.

If you have a 5 disks per EMC available, I would suggest, do not
create one LUN on this five disk, but rather each disk to be one LUN,
and group this 5 disk per EMC in one failure group...and let the ASM
do the striping and mirroring with another box (second failure group).

Even if a lot a people think that ASM is a "black box", it is a great
tool...and if you know the way how it works, and configure it
correctly, you can make your life much easier.


Mit freundlichen Grüßen,

Goran

------------------------------

Subject: RAC and ASM disk layout
Date: Fri, 9 Jun 2006 12:41:43 +0200
From: "Daiminger, Helmut" <HELMUT.DAIMINGER@xxxxxx>

Hi!

We are new to RAC and are discovering serveral uncertainties regarding
the best disk layout to use for ASM. We are running two EMC Clariion
systems (under RAID-5) and want ASM to write to both Clariions
simultaneously (i.e. use normal redundancy). ASM will balance the load
across the two failover groups (i.e. two LUNs) which reside in two EMC
Clariion arrays.

The question is, whether it is better to use a single LUN with 100 GB or
have a number of smaller LUNs (totalling 100 GB). Since we are using a
single raid group (i.e. 5 disks) for all database storage, all LUNs will
end up on the same physical devices anyway. So it might be harmful
having ASM to try and balance I/O across several LUNs which all reside
on the same physical devices.

This is 10.2 on HP-UX 11.23.

Thanks,
Helmut
--
//www.freelists.org/webpage/oracle-l


Other related posts: