Re: Logical IO

  • From: John Glohahi <john.o.golden@xxxxxxxxx>
  • To: Kellyn Pedersen <kjped1313@xxxxxxxxx>
  • Date: Thu, 19 Nov 2009 08:50:14 +0800

I ever tried to trace such CR events as 10200, but found they did not log
the reads on some special blocks, such as segment header. It can not
calculate the exact same reads number from the events trace files as the
number reported in the autotrace stat data.

Rgds,
John

On Wed, Nov 18, 2009 at 11:25 PM, Kellyn Pedersen <kjped1313@xxxxxxxxx>wrote:

>   Sheesh Dan, you could have warned us how many files were in that zip! :)
> Good stuff though!
>
> Kellyn Pedersen
> Multi-Platform DBA
> I-Behavior Inc.
> http://www.linkedin.com/in/kellynpedersen
>
> "Go away before I replace you with a very small and efficient shell
> script..."
>
>
> --- On *Wed, 11/18/09, Daniel Fink <daniel.fink@xxxxxxxxxxxxxx>* wrote:
>
>
> From: Daniel Fink <daniel.fink@xxxxxxxxxxxxxx>
> Subject: Re: Logical IO
> To: ukja.dion@xxxxxxxxx
> Cc: john.o.golden@xxxxxxxxx, oracle-l@xxxxxxxxxxxxx
> Date: Wednesday, November 18, 2009, 5:54 AM
>
>   A few years ago, Kirti Deshpande and I did some research into tracing
> Logical I/Os. The paper and presentation at are
> http://www.optimaldba.com/papers/TracingLogicalIOs.zip
>
>
>
> Dion Cho wrote:
> > Wouldn't it be better to trace it yourself with 10200 event and/or 10046
> event?
> >
> > - 10200 event would tell you which blocks are logically read.
> > - 10046 event would tell you which blocks are physicall read(when the
> buffer cache was clean).
> >
> > The number of blocks being read would also be depedent on how the segment
> space is managed(auto vs. manual) , which would be easily verifed with 10046
> and/or 10200 event.
> >
> >
> > ================================
> > Dion Cho - Oracle Performance Storyteller
> >
> > http://dioncho.wordpress.com (english)
> > http://ukja.tistory.com (korean)
> > http://dioncho.blogspot.com (japanese)
> > http://ask.ex-em.com (q&a)
> > ================================
>
>
> --
> //www.freelists.org/webpage/oracle-l
>
>
>
>

Other related posts: