Re: Surprising Performance Changes with Oracle (Long Post)

From: Mladen Gogala <>
Date: Thu, 10 Sep 2009 11:20:48 +0000 (UTC)
Message-ID: <h8aneg$lsi$>

Na Thu, 10 Sep 2009 02:20:49 -0700, Gerard H. Pille napisao:

> On 10 sep, 10:26, Mladen Gogala wrote:

>> Question: was direct I/O enabled here? If not, the blocks may have been
>> in the system buffer cache in which case oracle would still report them
>> as "physical reads" but blocks would be coming from memory instead.

> Why do you ask this? Charles clearly states that it was not. Will you
> believe him next time??

Because this was the new execution and I lost track of the other conditions. If direct I/O was disabled, Linux buffer cache would have contained the cached blocks and that would explain the difference in performance.

Received on Thu Sep 10 2009 - 06:20:48 CDT

Original text of this message