Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Oracle and disk striping
From: "MotoX" <rat_at_tat.a-tat.com> >>
said in answer to question on stripping >>
If you have small, random reads, then no, striping those tables is not a good
idea - the overhead of syncing the disks will make things slower. <<
I disagree since the whole point of stripping is to try to randomize all reads/writes to make every I/O an average I/O. Stripping works well with OLTP systems since it even outs the disk location of the requests and eliminates hot disks. This far and away exceeds the sync cost.
It also works with data wharehouses because the randomizing effect of the stripping is generally better than the result obtained by manual placement. Even if you are good and place the table extents wisely they will over time become unbalanced. OS level stripping stayed balanced better.
However the width of the stripe accross the disks is important to the
performance of both types of systems. For high sequential read of large data
areas wider widths work better.
While smaller widths of 32k or 64k work best for the OLTP systems.
These comments are based on personal experience and a book written by a senior
performance and tuning specialist with Oracle who had Oracle test results to
back him up.
The book is at work so I am not sure of the title, but I think it was Unix and
Oracle Performance Tuning. The publisher is Prentence Hall.
Mark Powell -- Oracle 7 Certified DBA
- The only advice that counts is the advice that you follow so follow your own
advice -
Received on Thu Sep 10 1998 - 17:56:44 CDT