Re: ASSM in 10g RAC doesnt seem work that well
- From: Christo Kutrovsky <kutrovsky.oracle@xxxxxxxxx>
- To: zoran_martic@xxxxxxxxx
- Date: Wed, 30 Mar 2005 16:54:28 -0500
I am still interested in this.
Martic, sure lets' ignore the waits, but real times speak for
themselves. +50% in execution time.
And the fact that I've confirmed that L1 block ownership is all from
the same instance is very strange. Almost smells like a bug. I can't
even begin to imagine the nightmare of trying to explain this to
Oracle.
--
Christo Kutrovsky
Database/System Administrator
The Pythian Group
The Pythian Group
On Tue, 29 Mar 2005 10:02:23 -0800 (PST), Martic Zoran
<zoran_martic@xxxxxxxxx> wrote:
> Christo,
>
> I could not reproduce the same CPU time for ASM and
> non-ASM tablespaces in all mine environments.
> I have only Solaris/HP-UX environments and so cannot
> test Linux.
>
> It could be OS specific of course. This is why I
> mentioned many times my environment and what I tested.
>
> I am not using any specific parameter.
> Maybe tuned a little bit interconnection between nodes
> in the past and that is all.
>
> In your case you need to tune waits because that is
> what is the different in your case.
> Nothing to do with the CPU usage because it is the
> same.
> I assume your test is far from the real scenario?
>
> Maybe you just proved when pushing with very high
> speed DML from both nodes then the collisions on
> blocks are more frequent and so pushing RAC to the end
> where performances starting to drop because of global
> cache.
> This is very natural. As faster you push both sides of
> RAC you are going to pay the price.
> In this particular case old non-ASSM with two
> freelists gave you better results.
>
> I assume your HW/SW/interconnect environment may cause
> these things. You should be aware about the RAC tuning
> and global cache issues with bad interconnection.
>
> Regards,
> Zoran
>
> --- Christo Kutrovsky <kutrovsky.oracle@xxxxxxxxx>
> wrote:
>
> > So right now we're basically lost?
> >
> > - We dont know why ASSM is using less CPU in
> > Martic's case and not in mine
> > - We dont know why ASSM is not working as expected
> > in my RAC test
> > case, and not in his.
> >
> > Any ideas ?
> >
> > Martic, why dont you post all your non-default
> > parameters ? Could it
> > be something OS specific, i did not see any LINUX
> > results or I missed
> > it ?
> >
> >
>
> __________________________________
> Do you Yahoo!?
> Yahoo! Small Business - Try our new resources site!
> http://smallbusiness.yahoo.com/resources/
> --
> //www.freelists.org/webpage/oracle-l
>
--
Christo Kutrovsky
Database/System Administrator
The Pythian Group
--
//www.freelists.org/webpage/oracle-l
Other related posts:
