Re: db_file_multiblock_read_count
From: Keith Moore <kmoore_at_zephyrus.com>
Date: Mon, 29 Sep 2008 20:25:15 -0500 (CDT)
Message-ID: <a2678c5361e44f092ca9a1be952f5839.squirrel@lady.zephyrus.com>
Date: Mon, 29 Sep 2008 20:25:15 -0500 (CDT)
Message-ID: <a2678c5361e44f092ca9a1be952f5839.squirrel@lady.zephyrus.com>
The writes will not cross extent boundaries so if you use AUTOALLOCATE for
extent management and it starts with 64K extents then that is all that will be
returned in a single read. Once the extent size reaches 1M or larger then it's
not an issue.
Keith
> I have always been a little humored by this parameter.
> db_file_multiblock_read_count attempts to specify a maximum value, as does
> the MBRC collected by system stats. However, that does not determine how
> many blocks are actually scooped up in a multiblock pass. One good example
> is to set event 10046 on a full table scan with waits (level 8 or 12) and
> look for db file scattered read waits, which will have the number of blocks
> read during the wait.
-- http://www.freelists.org/webpage/oracle-lReceived on Mon Sep 29 2008 - 20:25:15 CDT