Re: Partition LUNs for Oracle ASM
From: Radoulov, Dimitre <cichomitiko_at_gmail.com>
Date: Sun, 27 Jan 2013 11:00:19 +0100
Message-ID: <5104FAB3.6020804_at_gmail.com>
On 26/01/2013 18:58, Yong Huang wrote:
[...]
> Your mention of partition alignment is interesting. From Sun ZFS
> Storage Appliance: Performance clues and considerations [ID 1213714.1]
> we can infer that partition misalignment has performance implication,
> at least for Sun ZFS, and the first referenced article "Aligning
> Partitions to Maximize Storage Performance" goes to the details. So
> maybe we can say that not considering any bug, both not creating a
> partition and creating a partition but making sure it's aligned with
> the file system are acceptable, while creating a partition without
> proper alignment won't cause any problem other than performance
> degradation. What do you think?
Date: Sun, 27 Jan 2013 11:00:19 +0100
Message-ID: <5104FAB3.6020804_at_gmail.com>
On 26/01/2013 18:58, Yong Huang wrote:
[...]
> Your mention of partition alignment is interesting. From Sun ZFS
> Storage Appliance: Performance clues and considerations [ID 1213714.1]
> we can infer that partition misalignment has performance implication,
> at least for Sun ZFS, and the first referenced article "Aligning
> Partitions to Maximize Storage Performance" goes to the details. So
> maybe we can say that not considering any bug, both not creating a
> partition and creating a partition but making sure it's aligned with
> the file system are acceptable, while creating a partition without
> proper alignment won't cause any problem other than performance
> degradation. What do you think?
Hi Yong,
I agree. Yes, as far as I know, the only implication of the partition
misalignment is the performance impact:
2 reads on the storage side for a single read on the client side.
Considering that I couldn't foresee all possible effects of using
non-partitioned disk units
and that Oracle suggests to use disk partitions, I prefer to partition
the LUNs for now.
Thanks for the references: I didn't read note 1213714.1.
Regards
Dimitre
-- http://www.freelists.org/webpage/oracle-lReceived on Sun Jan 27 2013 - 11:00:19 CET