Oracle-L: T500 striping for performan

From: Steve Lemme <Steve_Lemme_at_CHDQM2.SPS.MOT.COM>
Date: Fri, 26 Jan 1996 15:41:10 -0500
Message-Id: <9601262227.AA17637@alice.jcc.com>

We just recently moved to a T500. We stripped at the minium of 1MB stripes, we would have preferred 64K instead but the limitation was 1Mb. The striping went across multiple disks depending upon the schema, and the partitions were sized according to the expected datafile sizes.
>From there, I striped multiple database files across multiple disks and I/O
channels for individual tablespaces for better performance. Some of our Tablespaces are 10 + Gb in size, therefore no data file was larger than 500Mb.

In this case we use ORACLE striping and HP striping but not RAID level 5 as we have seen performance impacts there.

After all was done we have seen a minimum of 10x increase in performance. ( you mileage may vary).

I hope to present at one of this years ORACLE conference our experiences.

Steve Lemme
DBA/Systems Manager
Motorola MOS-12

Date: 1/26/96 1:05 PM

> janicekp_at_aol.com (JaniceKP) wrote:
>
> :We have a T-500 that we wish to stripe for performance w/ I/O. Has anyone
> :done this before, and when creating the volume groups, what is the optimum
> :number of logical volumes per volume group?
>
> :This is a little new to us and we know we need to do this by hand as we
> :are being very specific in our mapping as to where we want to place
> :specific logical volumes, can anyone give me an easy step through set of
> :commands to do this when referring to the disk itself? I know I have them
> :in one of my notebooks from class, just figured I'd ask first.

Note that if your machine is used to run a database server, you may be better to use the database's striping instead. We run Oracle on a K100, soon to be a T500. Striping Oracle is more efficient and reliable (in terms of tablepsace recovery) that striping the FS. Striping BOTH Oracle and HPUX can degrade performance.

* * * * * * * * * * * * * *   dave
* DSylva_at_colonial.com.au  *
* * * * * * * * * * * * * *   std disclaimers apply...

Received on Fri Jan 26 1996 - 17:27:33 CST