Re: Getting the best out of new server hardware: a disk setup for Oracle database

On Tue, 13 Jun 2006 19:01:35 +0300, Heikki Siltala wrote:

> No RAC here. I found ASM to be the easiest way to implement S.A.M.E on
> this setup where we have six independent RAID 1+0 sets directly
> connected to the server. ASM will stripe everything over the six sets
> and quarantee that there are no "hot spots". This sounds much simpler
> than putting ext3 on the RAID sets and then try to manually distribute
> the IO load over the sets by manually relocating Oracle datafiles
> between the sets.

Let me dispel the myth of re-balancing by an email message from Scott Heisey, a guy who might know a thing or two about RAC, ASM and such. The message is from another Oracle forum:

Subject:	Re: ASM Record Deletions
Date:	06/13/2006 07:26:17 PM
From:	Scott <oraracdba_at_yahoo.com>
To:	smishra_97_at_yahoo.com, oracle-l_at_freelists.org

Sanjay, ASM works at the file level not at the table level. If you delete records the space in ASM does not change so there will be no re-balancing. Re-balancing will or should occur if you add, expand or drop tablespaces/datafiles.

Scott

• Sanjay Mishra <smishra_97_at_yahoo.com> wrote:

> Hi
>
> If I have 10 millions records in a table and if I
> drop the table or delete 1 million records and if
> ASM_POWER_LIMIT to > 0, Is it going to rebalance the
> Disk space among all Disks used in the Diskgroup.
> This table is currently having 50G space.
>
> TIA
> Sanjay
>

Both Sanjay Mishra and Scott Heisey are extremely well known names, they are no dilettantes. From that perspective, I/O load balancing is not a particularly useful feature, unless you plan on dropping your files daily, which would be a very novel approach to database administration. My advice to you is to contact everybody who's using ASM in production and ask them for experience. I expect you to find a s*tload of people using ASM in production, so you will have no problem finding references.

-- 
http://www.mgogala.com