Re: deciding which query plan to use
- From: Stephane Faroult <sfaroult@xxxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Thu, 15 Apr 2004 00:04:28 +0200
Ryan,
Somebody, I think it is Jocke Treugut, made you an excellent answer,
enquiring about the elapsed time. Logical I/Os are indeed a usually
excellent measure of performance, but they are not necessarily the final
answer. I have seen a number of cases (especially with IOTs) similar to
yours, much fewer LIOs and nevertheless longer to execute. I remember an
excellent case were the faster query was, consistently on several runs,
not the one with the best statistics (although it was far from being the
worst one). Latches, concurrency, many things may intervene. I
understand quite well that what you want to do is understand why, first
of all; but I doubt that in all cases all answers are easily available.
I don't think I have ever wasted my time trying to reduce LIOs. But at
some point, you indeed sometimes find that the best statistics don't
always give the fastest query, not more than the place you love best on
earth is necessarily the one with the best 'quality of life' statistics
(even if the correlation is usually high). Since my customers pay me to
make their programs run faster, not to show them the best statistics, my
ultimate yardstick is my wrist-watch :).
My 0.02 EUR
SF
Ryan wrote:
>
> yeah, but this is logical I/Os? That is all CPU isnt it?
> ----- Original Message -----
> From: "Mladen Gogala" <mladen@xxxxxxxxxxxxxxx>
> To: <oracle-l@xxxxxxxxxxxxx>
> Sent: Wednesday, April 14, 2004 9:04 AM
> Subject: Re: deciding which query plan to use
>
> >
> > On 04/14/2004 08:04:59 AM, ryan.gaffuri@xxxxxxx wrote:
> >
> > > Anyone ever run into this situation? Where logical I/Os are radically
> reduced, but overall stress on the system > > is radically increased?
> >
> > Well, you have to use something on your box. The thing is that you can't
> use disks and CPU
> > simultaneously in most of the cases. So, if you cut down the number of
> I/O's, you'll use
> > more CPU. It's sort of a natural thing, as much as anything about
> computers can be natural.
> > I am expecting to see herbal computers with no fat or cholesterol any day
> now. Balancing
> > will be much better with next generation of computers, the one which will
> have the RPM
> > instruction implemented in the hardware.
> > --
> > Mladen Gogala
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------
Other related posts:
