Re: Performance Diagnosis

From: Matt <mccmx_at_hotmail.com>
Date: 21 Jun 2004 00:52:41 -0700
Message-ID: <cfee5bcf.0406202352.745c53a9@posting.google.com>

> Would these waits be happening while you are checkpointing? Have you
> checked out DB_BLOCK_MAX_DIRTY_TARGET/FAST_START_IO_TARGET?
> http://download-west.oracle.com/docs/cd/A87860_01/doc/server.817/a76992/chx_inst.htm#434312
> The general idea for your case is that you write out dirty buffers
> more often, which lowers the wait time because there are fewer to
> write at a given time. Of course, if you are generating many dirty
> buffers, this might not work.
>
> Which version and platform, anyways?
>
> jg

8.1.7.4.1 on Windows...

DB_BLOCK_MAX_DIRTY_TARGET/FAST_START_IO_TARGET are both set to 150000 (default value).

I have set log_checkpoint_timeout to 0, and interval to 100000 in order to force checkpoints to occur at log switches...

My redo logs are 250Mb, however during the batch period the logs can switch nearly every 2 mins so we are generating alot of dirty buffers.

Also my buffer cache is very large... (over 1.2Gb)

There is also alot of truncate activity going on so this can force more dirty buffers to get written out.

Matt Received on Mon Jun 21 2004 - 02:52:41 CDT