Re: Need help: wio rease from 1% to more than 50% after increasing SGA size

From: joel garry <joel-garry_at_home.com>
Date: Fri, 27 Jun 2008 11:24:22 -0700 (PDT)
Message-ID: <903c208a-77ee-4f29-a903-0002caea8a75@m73g2000hsh.googlegroups.com>


On Jun 27, 7:36 am, "jshen...._at_gmail.com" <jshen...._at_gmail.com> wrote:
> > I don't think 8.1.7 is even supported any more.  Upgrade to something a
> > little more recent.
> > To find out why trace what the queries are doing.  What is taking so long?
>
> maybe that's a solution, but currently it could not help me out of
> trouble.
>
> To my understanding,  enlarge data buffer will do help with oracle
> performance, but whil wio increase so fast?

Look on metalink for notes about async i/o and hp-ux. If you aren't using raw devices, you may be seeing the effects of Oracle trying to be cute and using async i/o when you don't have it.

If you are using raw devices, you may be seeing the effect of moving bottlenecks around, like squeezing a balloon animal. You can see odd paradoxical effects when you just throw memory at Oracle. Sometimes it's things like, you do more stuff in memory, therefore you generate more redo, therefore you create an I/O problem.

jg

--
@home.com is bogus.
'Indeed, FYO point may ultimately be renamed the "FYRH point."' -
Bryan Cantrill, August 2004
Received on Fri Jun 27 2008 - 13:24:22 CDT

Original text of this message