Re: Does Oracle server try to allocate extents on five-block boundaries?

On Sat, 14 Jun 2003 17:54:55 +0300, "Sergey Adamenko" <adamenko_no__s_p_a_m_at_i.com.ua> wrote:

>Hi!
>
>I'm reading "Performance and Tuning" Student Guide for 8i.
>
>In chapter 9 "Using Oracle Blocks Efficiently" said:
>
>| Larger extents can improve performance slightly because the Oracle
>| server can read one large extent from disk with fewer multiblock reads
>| than are required to read many small extents. To avoid partial multiblock
>| reads, set the extent size to a multiple of 5 *DB_FILE_MULTIBLOCK_READ_COUNT.
>| Multiply by five because the Oracle server tries to allocate extents on five-block boundaries.
>| By matching extent sizes to the I/O and space allocation sizes, the
>| performance cost of having many extents in a segment is minimized.
>| However, for a table that never has a full table scan operation, it makes no
>| difference in terms of query performance whether the table has one extent
>| or multiple extents.
>
>I woudered about mentioned algorithm of "...allocating extents on five-
>block boundaries". No mention about it in "Oracle Documentation
>Library, Release 8.1.7". Only good known principles of allocation for
>DMT with its "MINIMUM EXTENT" and LMT with its "UNIFORM SIZE" or
>"AUTOALLOCATE".
>Same results when searching in Goggles.
>
>My question is: is this principle discarded relating to 8.1.7 or I missing something?
>
>Thank you!
>
>Sergey Adamenko.
>

It is not discarded. If you don't specify a minimum extent in a DMT your extents will still be rounded up to a multiple of 5. The idea probably originated in the days when a 2k blocksize was the norm: one header block and 4 datablocks.

Sybrand Bakker, Senior Oracle DBA

To reply remove -verwijderdit from my e-mail address Received on Sat Jun 14 2003 - 12:57:48 CDT