| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> RE: Why a rebuild speeds up my queries.
Dave,
I'm of course nervous basing a diagnostic decision on a statspack =
report.
Your extended SQL trace data would give you enough information to =
determine
whether the extra PIO blocks are caused by migrated rows, and if so, =
exactly
how much time you should expect to save by fixing the migrated rows =
problem.
Assuming that your problem is a bad PCTFREE setting, now... A good =
sanity
check should be easy. You know how big the rows start out. You know how =
big
the rows end up (you can measure that). And you know the definition of
PCTFREE (from the Concepts guide). You ought to be able to compute a =
PCTFREE
number that greatly reduces your pain from its present level, even if =
the
value you choose isn't perfect (either wastes a little space, or still
produces a few migrated rows).
Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com
* Nullius in verba *
Upcoming events:
- Performance Diagnosis 101: 9/14 San Francisco, 10/5 Charlotte, 10/26
Toronto
- SQL Optimization 101: 9/20 Hartford, 10/18 New Orleans - Hotsos Symposium 2005: March 6-10 Dallas - Visit www.hotsos.com for schedule details...
-----Original Message-----
From: oracle-l-bounce_at_freelists.org =
[mailto:oracle-l-bounce_at_freelists.org]
On Behalf Of David Sharples
Sent: Wednesday, September 01, 2004 7:48 AM
To: oracle-l_at_freelists.org
Subject: RE: Why a rebuild speeds up my queries.
Hi Richard thanks for the snippet about dbms_stats, yes we do use that. Next time I will use analyze and see what that is.
As for pctfree, the tables are set to the default of 10. I think this is the problem because the rows start of as 'skinny' ones with all values being 0, then they get updated constantly with true values, numbers up to 10 digits, varchar2 up to 20, so they then become fat and must migrate.
Do you know of a good link where it tells me how to set pctfree properly?
Oracle docs are good but I find them to tell me more what it is about that what it should be.
Thanks very much
Dave
-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Richard Foote
Sent: 01 September 2004 13:22
To: oracle-l_at_freelists.org
Subject: Re: Why a rebuild speeds up my queries.
Hi David,
Your "table fetch continued row" count has certainly plummeted so something's changed there.
When you say dba_tables shows nothing, you don't by any chance use
dbms_stats to analyze as chain_cnt is only calculated with the "old"
analyze
command (else you just get a 0).
In which case, yes, your pctfree is buggered.
Just a thought.
Cheers
Richard
----- Original Message -----
From: "David Sharples" <dsharples_at_cerebrussolutions.com>
To: <oracle-l_at_freelists.org>
Sent: Wednesday, September 01, 2004 9:21 PM
Subject: Why a rebuild speeds up my queries.
Hi,
I have a process that overtime slowly gets slower and slower (execution
plans are the same)
A rebuild of the table / index fixes this and makes it go quick again. I know that we shouldn't need to rebuild things, so I need to know what is wrong in my setup which is causing this.
The setup is this: oracle 9.2.0.4 on Solaris
Running queries against hashed partitioned table which never get deleted them, they only get inserted into and then updated a fair amount.
We think it is due to row migration / chained rows but chain count from dba_tables showed nothing after an analyze.
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.htmlput 'unsubscribe' in the subject line.
-----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request_at_freelists.org
 |
 |