Re: HELP: Info about ORACLE7 Performance on RAID

From: <Will>
Date: 1995/08/08
Message-ID: <406gd8$sui_at_data.interserv.net>#1/1


> Peter Constantinou <rosepete_at_xs4all.nl> writes:
> X-URL: news:comp.periphs.scsi
> Content-Transfer-Encoding: 7bit
> Content-Type: text/plain; charset=us-ascii
>
> Does any one have experience of ORACLE7 applications running on RAID disk
> arrays. I have been told that performance is poor for Oracle V6 applications,
> but cannot find anyone with experience of ORACLE7.
>
> FYI The application is ORACLE7 v7.0 running on IBM RS/6000 under AIX 3.2.5
> (soon to be ORACLE7 v7.1 under AIX v4.0). We are looking to purchase 12Gb of
> Seagate or IBM based RAID (configured for RAID level 5).
>
> Thanks in advance for any info
>
> Peter C
>
> --
> rosepete_at_xs4all.nl Phone: +31 10 4987 727 Fax: +31 10 2120 295
> EDS BV, Electronic Data Systems, The Netherlands.
>
>

>>>>

I've used RAID 0,1,0+1, and 5 with Oracle 6 and 7, and saw no difference in performance between the two.

I didn't see a lot of improvement in TP oriented access, but I saw a big improvement in decision support queries (table scans).

I'm not a big fan of RAID 5 due to it's VERY poor write performance. If you think it's bad when all the disk drives are working, wait until a drive fails. You still have your array, but the system seems to halt due to poor performance.

My absolute favorite is RAID 0+1. This gives you redundancy, with unbelievable performance. But, it costs a lot since you must buy twice as many disk drives as you would have to otherwise.

Quick RAID overview:

Without going into a lot of detail, here's the scoop on the most popular RAID levels.

RAID 0 (mirroring)	reads are slightly faster,  writes are slightly slower  (with redundancy)
RAID 1 (striping)	reads are much faster, writes are much faster  (no redundancy)
RAID 0+1		reads are much faster, writes are much faster,  (with redundancy)
RAID 5		reads are much faster, writes are much slower (with redundancy)


Hope this helps,
Will Kooiman Received on Tue Aug 08 1995 - 00:00:00 CEST

Original text of this message