Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> RE: frequency + height based histograms.

RE: frequency + height based histograms.

From: Christian Antognini <Christian.Antognini_at_trivadis.com>
Date: Mon, 6 Mar 2006 12:55:24 +0100
Message-ID: <F2C9CCA71510B442AF71446CAE8AEBAF3F5476@MSXVS04.trivadis.com>


Jaffar

>I was reading Jonathan's great book, at least for me, 'CBO
>Fundamentals' and I got confused of frequency/height based histograms,
>though he explained in a great way.

What is confusing is the name given by Oracle ;-)

>What I understood is that, if we have more than 254 distinct values in
>a columns, we use height based histograms and when we have less than
>254 distinct values in a column, we use one bucket for 1 distinct
>value. Are my assumption are correct, or am I missing something?

if
  "1 distinct value per bucket"
then
  "frequency"
else
  "height-based"
end if

i.e. you may have a height-based histogram even if there are "only" 50 or 100 distinct values. It depends on how you created it.

HTH
Chris

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Mar 06 2006 - 05:55:24 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US