Re: [Q] db_block_size: 8K vs 2K default: Why is it better?

From: Steve Adams <steve.adams_at_ixora.com.au>
Date: Mon, 14 Feb 2000 07:01:14 GMT
Message-ID: <38a7626e.610977709@news.eagles.bbs.net.au>

Hello Stephen,

The key factors are the I/O cost of index range scans, and the extra rotational latency for partial block writes. If you make extensive use of index range scans, or have a DBWn bottleneck, then the difference is likely to be very noticeable. Otherwise, it may not be. If you have not already seen it, there is a tip on this matter on my web site.

Regards,
Steve Adams

http://www.ixora.com.au/

http://www.oreilly.com/catalog/orinternals/

http://www.christianity.net.au/

On Sun, 13 Feb 2000 16:30:42 -0500, Stephen Hurrell <hurrells_at_hotmail.com> wrote:

><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
><html>
>OK.
><p>Still looking for valid information about the db_block_size effect.
><p>I have read various documents describing why you would adjust your block
>size from the default 2K.
><br>These articles, books and personal opinions break down as follows:
><ul>
><li>
>Changing from the default 2k has no noticeable affect.</li>
>
><li>
>Changing from the default 2k upto 4k or 8k has a big noticeable effect.</li>
>
><li>
>You should change oracle db_block_size to match your os or db size (e.g.:
>8k on HP-UX, 4k on Solaris 2.x)</li>
></ul>
>I understand the relationship between larger block size results in fewer
>block reads from disk, shallower indexes and all sorts of OS performance
>improvements.</html>
>
Received on Mon Feb 14 2000 - 01:01:14 CST