Re: Oracle 7.3.2 and RAID

From: Phil Herring <revdoc_at_uow.edu.au>
Date: 1997/10/12
Message-ID: <61rnv5$fh2$1@wyrm.its.uow.edu.au>#1/1

In article <61f5j5$q9h$1_at_gruvel.une.edu.au> Craig Munday, cmunday_at_neumann.une.edu.au writes:
>I was just wondering if anybody has had any experience with Oracle 7.3.2
>running on a RAID subsystem.

I'd recommend RAID-0+1 for all your data. RAID-5 is a bit cheaper than RAID-0+1, but the savings are usually small, and you might not get the same write performance. (OTOH, you might.)

Put *all* of your data on redundant storage of some kind - even if you only lose your redo logs, your system will still stop. The cost of one extra disk is usually trivial compared to the cost of an outage.

Stripe size is critical to performance. You may have to run some tests with your application to determine the best value. I've read that DB_BLOCK_SIZE * DB_FILE_MULTIBLOCK_READ_COUNT is a good starting point.

Lastly, your logical volumes should be on separate physical devices. In other words, don't have two volumes sharing disks or controllers. If you do this, you can simply treat each logical volume as if it were a single disk, in which case all of your data placement decisions are the same as they would be if you weren't using RAID.

Once you've striped your data, set up the Parallel Query Option. Oracle have a white paper on this on their support web site. You'll probably get a performance boost, even with one CPU.

Received on Sun Oct 12 1997 - 00:00:00 CDT