| 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: full-scan vs index for "small" tables
we are both right.
In the real life I had kind of
select where type in ('OPENED', 'ACTIVE')
Anyway, this does not change anything regarding the CBO darned undeterministic nature. I did a small test to reproduce my case. The results are more than funny. I really do not want to examine whatever traces. I see no sense in that. If it's bug then Oracle's not gonna pay me for finding it. If it is not then what for I have to dig? The Oracle version is not the last (9206) also I assume many people are still running v7.3. Here bellow is my test. Someone claimed CBO has a sense of humor? A black humor, I have to say.
Rem drop table test_cbo;
create table test_cbo (id number, type varchar2(64));
alter table test_cbo add constraint test_cbo_pk primary key (id);
create index test_cbo_type on test_cbo (type);
begin
for m in 1..9999 loop
insert into test_cbo values (m, 'CLOSED');
end loop;
commit;
end;
/
begin dbms_stats.gather_table_stats (user, 'TEST_CBO', method_opt=>'FOR ALL COLUMNS SIZE AUTO', cascade=>true); end; /
select * from TEST_CBO where type='CLOSED';
Statement Id=0 Type=SELECT STATEMENT Cost=8 TimeStamp=29-06-06::17::32:08
(1) SELECT STATEMENT CHOOSE
Est. Rows: 9.999 Cost: 8
(2) TABLE ACCESS FULL LTDLNE.TEST_CBO [Analyzed]
(2) Blocks: 28 Est. Rows: 9.999 of 9.999 Cost: 8
Tablespace: USERS
select * from TEST_CBO where type in ('OPENED' , 'OPENED2')
Statement Id=0 Type=SELECT STATEMENT Cost=8 TimeStamp=29-06-06::17::32:45
(1) SELECT STATEMENT CHOOSE
Est. Rows: 9.999 Cost: 8
(2) TABLE ACCESS FULL LTDLNE.TEST_CBO [Analyzed]
(2) Blocks: 28 Est. Rows: 9.999 of 9.999 Cost: 8
Tablespace: USERS
select * from TEST_CBO where type = 'OPENED'
union
select * from TEST_CBO where type = 'OPENED2'
Statement Id=2 Type=UNION-ALL
Cost=0 TimeStamp=29-06-06::17::33:08
(1) SELECT STATEMENT CHOOSE
Est. Rows: 10.000 Cost: 16
(6) SORT UNIQUE
Est. Rows: 10.000 Cost: 16
(5) UNION-ALL
(3) TABLE ACCESS BY INDEX ROWID LTDLNE.TEST_CBO [Analyzed]
(3) Blocks: 28 Est. Rows: 1 of 9.999 Cost: 3
Tablespace: USERS
(2) NON-UNIQUE INDEX RANGE SCAN LTDLNE.TEST_CBO_TYPE [Analyzed]
Est. Rows: 1 Cost: 2
(4) TABLE ACCESS FULL LTDLNE.TEST_CBO [Analyzed]
(4) Blocks: 28 Est. Rows: 9.999 of 9.999 Cost: 8
Tablespace: USERS
SQL Statement from editor:
select * from TEST_CBO where type = 'OPENED'
union
select * from TEST_CBO where type = 'OPENE2'
Statement Id=3 Type=TABLE ACCESS
Cost=3 TimeStamp=29-06-06::17::34:04
(1) SELECT STATEMENT CHOOSE
Est. Rows: 2 Cost: 8
(7) SORT UNIQUE
Est. Rows: 2 Cost: 8
(6) UNION-ALL
(3) TABLE ACCESS BY INDEX ROWID LTDLNE.TEST_CBO [Analyzed]
(3) Blocks: 28 Est. Rows: 1 of 9.999 Cost: 3
Tablespace: USERS
(2) NON-UNIQUE INDEX RANGE SCAN LTDLNE.TEST_CBO_TYPE [Analyzed]
Est. Rows: 1 Cost: 2
(5) TABLE ACCESS BY INDEX ROWID LTDLNE.TEST_CBO [Analyzed]
(5) Blocks: 28 Est. Rows: 1 of 9.999 Cost: 3
Tablespace: USERS
(4) NON-UNIQUE INDEX RANGE SCAN LTDLNE.TEST_CBO_TYPE [Analyzed]
Est. Rows: 1 Cost: 2
________________________________
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Cary Millsap
Sent: 29. júní 2006 16:54
To: oracle-l_at_freelists.org
Subject: RE: full-scan vs index for "small" tables
Exactly. That's a place where CBO should have no trouble whatsoever making the right decision (index access).
Cary Millsap
Hotsos Enterprises, Ltd.
Nullius in verba
Hotsos Symposium 2007 / March 4-8 / Dallas
Visit www.hotsos.com for curriculum and schedule details...
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Wolfgang Breitling
Sent: Thursday, June 29, 2006 6:44 AM
To: Laimutis.Nedzinskas_at_landsbanki.is
Cc: oracle-l_at_freelists.org
Subject: RE: full-scan vs index for "small" tables
At 03:20 AM 6/29/2006, Laimutis Nedzinskas wrote:
I can illustrate my point using one example (which seems to be one the the biggest CBO issues actually, discussed many times before):
I have a "select where type='OPENED' "
No need to say that type='OPENED' is 0.1% of all records, the rest are 'CLOSED'.
I built a histogram. Value 'OPENED' happened not to be there because... the were no records with type='OPENED' at the time. Plan was full scan of course.
That is not my experience at all. If the value in the predicate is NOT in the (frequency) histogram then in my experience hat is synonymous for "very rare" for the CBO and it uses the index. From all you are saying I get the impression that you are using a different CBO from the one I am using.
Regards
Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com
Fyrirvari/Disclaimer
http://www.landsbanki.is/disclaimer
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Jun 29 2006 - 12:38:33 CDT
 |
 |