Re: db_file_multiblock_read_count 10g default values

From: Greg Rahn <greg_at_structureddata.org>
Date: Thu, 5 Nov 2009 11:13:56 -0800
Message-ID: <a9c093440911051113y24e29d2ao69b7c9efb2540a74_at_mail.gmail.com>

The choice for db_file_multiblock_read_count when not set is dependent on some other parameters (IIRC buffer cache size, sessions/processes, likely some others) so the likely reason for the difference between instances is the difference in parameters.

Do you have some performance data that points to a larger IO size impacting performance or is this just an attempt to "optimize" without any performance data (by your comments seems to be the latter)? By collecting system stats you introduce the possibility to have plan change, which may have more downside risk than a few IOs that could be larger.

I'd say, if its not broke, it doesn't need fixing...

On Thu, Nov 5, 2009 at 9:57 AM, Allen, Brandon <Brandon.Allen_at_oneneck.com> wrote:

> I’ve got several Oracle 10.2.0.4 database instances running on the same
> Oracle Linux 5.2 box and from the documentation and everything else I could
> find, it looks like db_file_multiblock_read_count (dbfmbrc) is supposed to
> default to the maximum IO size of the OS (in units of db_block_size), but
> I’ve got varying sizes from 24 to 82 for dbfmbrc even though they’re all on
> the same host and have the same db_block_size (8k).  Any idea what could
> cause the differences?

-- 
Regards,
Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l

Received on Thu Nov 05 2009 - 13:13:56 CST