My thoughts - others may differ:

a) Only have histograms where absolutely necessary. 
If you've got large histograms on every table columns
thats a lot of stuff sucking up dictionary cache space
which will never get used.  Similarly, it would
probably be more effort to parse (although I have not
quantified that)

b) The number of distinct/duplicate values should
really be relevant - its whether you will regularly
need to probe a table using column values that are
skewed in such a way as to have the optimizer make
poor assumptions about their distribution.

c) You may want consider playing with the automatic
histogram collection in 9i, where a hidden table is
regularly populated with the columns and predicates
you've used and thus can make some decisions on
whether a histogram may benefit.

hth
connor

 --- Murali Menon <[EMAIL PROTECTED]> wrote: > 
> So far only had to use RBO, now we are moving
> towards using CBO.  However none of the documents
> really talk in detail about Histograms, the goods,
> bads where and where not to use.  
> 
> I have read that HISTOGRAMS are good for columns
> that are non unique and that have many duplicate
> values. While DSS systems have more occurances of
> these kind of columns, OLTP databases could also
> have certain columns like status codes  or state
> codes that have duplicate values.  
> 
> Is it OK to have histograms generated for all
> columns irrespective of the type of system the
> database supports?  and pros and cons?
> 
> Could some one point me to a good white paper or
> material that would discuss this?
> 
> TIA
> 
> Menon
> 
> 
> 
> ---------------------------------
> Do you Yahoo!?
> Yahoo! Mail Plus - Powerful. Affordable. Sign up now 

=====
Connor McDonald
web: http://www.oracledba.co.uk
web: http://www.oaktable.net
email: [EMAIL PROTECTED]

"GIVE a man a fish and he will eat for a day. But TEACH him how to fish, and...he will 
sit in a boat and drink beer all day"

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com
-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: =?iso-8859-1?q?Connor=20McDonald?=
  INET: [EMAIL PROTECTED]

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Reply via email to