Re: I/O and db_file_multiblock_read

From: Mladen Gogala <mgogala_at_verizon.net>
Date: Sun, 10 Dec 2006 13:57:46 -0500
Message-id: <1165777066l.3000l.0l@medo.noip.com>

On 12/09/2006 10:34:42 AM, Vlad Sadilovskiy wrote:
> I don't understand why people insist on gathering system statistics as
> Hemant noted, instead of answering the question why higher MBRC parameter
> resulted in longer run. MBRC from system statistics is used solely for CBO
> purposes. The actual reading is done in accordance
> with db_file_multiblock_read_count.
>

When you gather system statistics, the value of MBRC does not influence the optimizer any more. The biggest single problem with MBRC was that it used to make full table scans appear cheaper and its increase would slant your execution plans toward full table scan. Now, you can collect the optimizer statistics with MBRC=8 and increase it to 16 or 32 without the price to pay.

-- 
Mladen Gogala
http://www.mladen-gogala.com

--
http://www.freelists.org/webpage/oracle-l

Received on Sun Dec 10 2006 - 12:57:46 CST

Re: I/O and db_file_multiblock_read_count