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: Function-Based Index not working

RE: Function-Based Index not working

From: Cary Millsap <cary.millsap_at_hotsos.com>
Date: Thu, 05 Sep 2002 13:33:31 -0800
Message-ID: <F001.004C922C.20020905133331@fatcity.com>


Right.

The index one is easy to explain. It's one LIO for the index root (which is also the appropriate leaf block for such a small index). Then it's one more LIO for the data block. In addition to there being only 2 LIOs, note also that these LIOs are also very inexpensive LIOs. All LIOs do not cost the same. LIOs that seek straight to a desired row are cheaper than LIOs motivated by a full-table scan. Each LIO motivated by a full-table scan has to parse an entire Oracle block.

Why the full-table scan requires five LIOs is a guess. First, remember that no matter whether my guess is correct or not, you are seeing proof that the index access is cheaper than the full-table scan. Now, my guess is grounded in remembering what a full-table scan does: it reads up to the high-water mark. Well, the high-water mark moves in increments of _bump_high_water_mark_count (sp?), which by default is 5. (If it defaulted to 1, people would see lots more waits for "buffer busy waits" events.) Since "select * from onerow" is a full-table scan, it reads five blocks. If you really want to dig into it, restart your test instance and run the experiment again. In the "db_file_%_read" event p1 and p2 values, you can see exactly which blocks are getting read.

Fun stuff. Establishing the habit of tracing things instead of just blindly believing what the so-called "experts" say is, in my estimation, the *most* important critical success factor for an Oracle performance analyst.

Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com

Upcoming events:
- Hotsos Clinic, Oct 1-3 San Francisco, Oct 15-17 Dallas, Dec 9-11 Honolulu
- 2003 Hotsos Symposium on OracleR System Performance, Feb 9-12 Dallas - Next event: Miracle Database Forum, Sep 20-22 Middelfart Denmark

-----Original Message-----
Sent: Thursday, September 05, 2002 2:16 PM To: ORACLE-L_at_fatcity.com
Cc: cary.millsap_at_hotsos.com
Importance: High

Cary,

Two buffer gets for an Index Range scan vs. 5 buffer gets for a Full Table
Scan.

Why is that?

Jared

FTS
Buffer gets: 5
Time: < 1/100 second



PARSING IN CURSOR #3 len=21 dep=0 uid=19 oct=3 lid=19 tim=45692398 hv=2775880792 ad='5304bbac'
select * from onerow
END OF STMT
PARSE #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692398 EXEC #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692398 WAIT #3: nam='SQL*Net message to client' ela= 0 p1=1413697536 p2=1 p3=0 FETCH #3:c=0,e=0,p=0,cr=1,cu=4,mis=0,r=1,dep=0,og=4,tim=45692398 WAIT #3: nam='SQL*Net message from client' ela= 0 p1=1413697536 p2=1 p3=0
FETCH #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692398 WAIT #3: nam='SQL*Net message to client' ela= 0 p1=1413697536 p2=1 p3=0 WAIT #3: nam='SQL*Net message from client' ela= 0 p1=1413697536 p2=1 p3=0
STAT #3 id=1 cnt=1 pid=0 pos=0 obj=4651 op='TABLE ACCESS FULL ONEROW '

IRS
Buffer gets: 2
Time < 1/100 second



PARSING IN CURSOR #3 len=33 dep=0 uid=19 oct=3 lid=19 tim=45692399 hv=2516078141 ad='531a22b8'
select * from onerow where key=1
END OF STMT
PARSE #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692399 EXEC #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692399 WAIT #3: nam='SQL*Net message to client' ela= 0 p1=1413697536 p2=1 p3=0 FETCH #3:c=0,e=0,p=0,cr=2,cu=0,mis=0,r=1,dep=0,og=4,tim=45692399 WAIT #3: nam='SQL*Net message from client' ela= 0 p1=1413697536 p2=1 p3=0
FETCH #3:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=45692399 WAIT #3: nam='SQL*Net message to client' ela= 0 p1=1413697536 p2=1 p3=0 WAIT #3: nam='SQL*Net message from client' ela= 0 p1=1413697536 p2=1 p3=0
STAT #3 id=1 cnt=1 pid=0 pos=0 obj=4651 op='TABLE ACCESS BY INDEX ROWID ONEROW '
STAT #3 id=2 cnt=2 pid=1 pos=1 obj=4652 op='INDEX RANGE SCAN '
WAIT #3: nam='SQL*Net message to client' ela= 0 p1=1413697536 p2=1 p3=0
WAIT #3: nam='SQL*Net message from client' ela= 0 p1=1413697536 p2=1
p3=0

"Cary Millsap" <cary.millsap_at_hotsos.com> Sent by: root_at_fatcity.com
 09/05/2002 11:13 AM
 Please respond to ORACLE-L  

        To: Multiple recipients of list ORACLE-L <ORACLE-L_at_fatcity.com>

        cc: 
        Subject:        RE: Function-Based Index not working


Even when the high-water mark thing isn't a problem, it's sometimes more

efficient to read every row in a table through an index than via a full-table scan.  

If you're curious, try this. Create a table with two columns, "key" and "value", and insert one row with key=1, value='x'. Create an index on "key". Then?  

alter session set events '10046 trace name context forever, level 8';

select * from onerow;  /* just to make sure it's cached */
select * from onerow;
select * from onerow where key=1;  /* just to make sure it's cached */
select * from onerow where key=1;

exit;  

Now look at your trace data. You'll find that the full-table scan of this
table is both cheaper and faster through the index.  

The age-old advice from many SQL tuning "experts" is badly wrong when they
tell you never to index small tables. For applications that execute a lot
of small-table queries, the performance impact really adds up.  

Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com

Upcoming events:
- Hotsos Clinic, Oct 1?3 San Francisco, Oct 15?17 Dallas, Dec 9?11 Honolulu

- 2003 Hotsos Symposium on OracleR System Performance, Feb 9?12 Dallas
- Next event: Miracle Database Forum, Sep 20?22 Middlefart Denmark
-----Original Message-----

Sent: Thursday, September 05, 2002 12:19 PM To: Multiple recipients of list ORACLE-L  

Not necessarily... Cary's IOUG-A presentation covers this very well. One

scenario is where the high water mark is set artificially high, and there
are far more blocks allocated than actually contain data. In this case, a
FTS will be reading far too many empty blocks. -----Original Message-----
Sent: Thursday, September 05, 2002 10:19 AM To: Multiple recipients of list ORACLE-L Hello  

I think that the amount of records you read is also taken into account. If you run a query that selects ALL the records in the tables it is ALWAYS more efficient to do full table scan then to access by index.  

Yechiel Adar
Mehish
----- Original Message -----
To: Multiple recipients of list ORACLE-L Sent: Saturday, August 31, 2002 4:23 PM  

Hi All,  

Thanks a lot to you all. At last I got the function-based index working properly.
This is what I noticed :-
Have to alter session/system for :-
+ alter session set QUERY_REWRITE_ENABLED=TRUE;
+ alter session set QUERY_REWRITE_INTEGRITY=TRUSTED;
+ alter session set optimizer_mode=FIRST_ROWS;
 

And
+ can't use IS NULL & IS NOT NULL clause.
+ can't use Like operator.
 

Regards,
Marul.      

Hi Naveen,
Thanks a lot for the efforts you are putting in for me for such a simple

problem, but unfortunately, for me all the tips and tricks are not solving
the problem.
Now these are my current statistics :-  

+ alter session set QUERY_REWRITE_ENABLED=TRUE;
+ alter session set QUERY_REWRITE_INTEGRITY=TRUSTED;
+ alter session set optimizer_mode=FIRST_ROWS;
+ alter session set DB_FILE_MULTIBLOCK_READ_COUNT=1;
 

This procedure writes 180,000 records in employeees table
+ execute bulk_insert
 

Analyzing table and rebuilding index (though its not necessary)
+ analyze table employees compute statistics;
+ alter index upper_ix rebuild;

Making autotrace on
+ set autotrace traceonly explain
 

Fired the query:
SELECT last_name FROM employees WHERE UPPER(last_name) IS NOT NULL ORDER
BY UPPER(last_name);
Elapsed: 00:00:00.00  

Execution Plan


   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=57 Card=4001 Bytes=2
          0005)
 
   1    0   SORT (ORDER BY) (Cost=57 Card=4001 Bytes=20005)
   2    1     TABLE ACCESS (FULL) OF 'EMPLOYEES' (Cost=38 Card=4001 By
          tes=20005)
 

Any clues what is happening? Should I insert more records in the table.  

TIA,
Marul.            

See the table's size is very small. Till it atleast 2 times the value of

DB_BLOCK_SIZE * DB_FILE_MULTIBLOCK_READ_COUNT it will not use index.  

Set the value of DB_FILE_MULTIBLOCK_READ_COUNT to one.  

Insert lots of values in the table. You can make a procedure to insert random characters into the table, and then put it in a big loop. Analyze

table and thn run the same query.  

It should work  

naveen

 -----Original Message-----
Sent: Saturday, August 31, 2002 4:03 PM
To: Multiple recipients of list ORACLE-L Thanks a lot Naveen,  

Even after executing the following the execution plan shows full table scan :-  

+ alter session set QUERY_REWRITE_ENABLED=TRUE;
+ alter session set QUERY_REWRITE_INTEGRITY=TRUSTED;
+ alter session set optimizer_mode=FIRST_ROWS;
+ Insert into employees values('A');
+ Insert into employees values('B');
+ analyze table employees compute statistics;
+

    select last_name
   FROM employees WHERE UPPER(last_name) IS NOT NULL    ORDER BY UPPER(last_name); 2 3
Elapsed: 00:00:00.00  

Execution Plan


   0      SELECT STATEMENT Optimizer=FIRST_ROWS (Cost=3 Card=2 Bytes=2
          )
 
   1    0   SORT (ORDER BY) (Cost=3 Card=2 Bytes=2)
   2    1     TABLE ACCESS (FULL) OF 'EMPLOYEES' (Cost=1 Card=2 Bytes=
          2)
 

Even after using the hint no change in the plan :-
+ select /* INDEX employees(upper_ix) */ last_name FROM employees WHERE
UPPER(last_name) IS NOT NULL;  

Please tell me what else should I do to make this query use the index which is created.    

TIA,
Marul.  

Marul,  

  1. you don't have table analyzed in which case Rule based optimizer will

be used. CBO is used if atleast one of the tables in the query is ANALYZED
2. There is no data in your table. Optimizer goes for a full tablescan if
it thinks that it will be moer advisable to do a full table scan. e.g. You
will not use the INDEX if your book has only one page.  

The decision of going for a full tablescan is based on DB_BLOCK_SIZE * DB_FILE_MULTI_BLOCK_READCOUNT, which tells how much data Oracle fetches at
one time. If your entire table can be fetched in atleast 2 fetches, full

table scan will be done instead of INDEX scan, to avoid doubling of work.    

Naveen
-----Original Message-----
Sent: Saturday, August 31, 2002 2:18 PM
To: Multiple recipients of list ORACLE-L Hi,  

Can you please help me out in solving this weird problem of funcation-based index not being used when I query the table. This is the comand I fired and the result it returned me.  

  1. SQL> create table employees (last_name varchar2(20)); Table created.
  2. SQL> CREATE INDEX upper_ix ON employees (UPPER(last_name)); Index created.

Made the autotrace on and than:-  

3. SELECT last_name FROM employees WHERE UPPER(last_name) IS NOT NULL ORDER BY UPPER(last_name);

    no rows selected.  

Execution Plan


   0 SELECT STATEMENT Optimizer=CHOOSE    1 0 SORT (ORDER BY)
   2 1 TABLE ACCESS (FULL) OF 'EMPLOYEES'     I fired without order by clause also but no use.  

Now can any body please let tell me why this Oracle is having a full scan
of the employee table.  

TIA,
Marul.        

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Cary Millsap
  INET: cary.millsap_at_hotsos.com

Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (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).
Received on Thu Sep 05 2002 - 16:33:31 CDT

Original text of this message

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