Performance issue in this query:

Question

Hi,
Here is the query:
SELECT COLLECTIONKEY, FLOWSYSKEY, LIFECYCLENUMBER, PUBLICVERSION, PUBLICOPERATION,
STATUSCODE, VALUEDATE, REASONCODE FROM AA out WHERE COLLECTIONKEY
LIKE '1437023L%' AND COLLECTIONVERSION = ( SELECT MAX(COLLECTIONVERSION) FROM cfdg_owner.AA inn WHERE inn.FLOWSYSKEY=out.FLOWSYSKEY)
Now this table AA has non clustered index on CollectionKey and CollectionVersion.
The above query will return only 4 records, but still the optimizer is using index .
Fyi, AA table is huge table with records in millions.
SELECT STATEMENT     8.0     8     58591     1     108     8                         ALL_ROWS                                                       
FILTER                              1                                                                                
TABLE ACCESS (BY INDEX ROWID)     5.0     5     36827     2     216     1     CFDG_OWNER     TB_PUBLIC_CF     BY INDEX ROWID     TABLE     ANALYZED     1                                                  
INDEX (RANGE SCAN)     3.0     3     21764     2          1     CFDG_OWNER     IDX2_TB_PUBLIC_CF     RANGE SCAN     INDEX     ANALYZED                    1                                   
SORT (AGGREGATE)                    1     41     2               AGGREGATE                                                                 
INDEX (RANGE SCAN)     3.0     3     21764     1     41     1     CFDG_OWNER     IDX1_TB_PUBLIC_CF     RANGE SCAN     INDEX     ANALYZED                    1
Plus the response time of this query is very fast. When i am forcing optimizer not to use Index, its taking more time.
How come Index range scan is performing better in this case as compared to FULL Scan?
Version is:
BANNER
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
PL/SQL Release 10.2.0.4.0 - Production
CORE     10.2.0.4.0     Production
TNS for Solaris: Version 10.2.0.4.0 - Production
NLSRTL Version 10.2.0.4.0 - Production
Any more information required, please let me know.
Thanks.
                                                  1

Answer 1

It only returns 4 records, but how many does it have to scan through and eliminate to satisfy the MAX condition? ie) how many versions are there per key?
It's quite possible this would be a more efficient query for you (it will obviate the secondary index range scan the current query is utilizing).
SELECT
  COLLECTIONKEY,
  FLOWSYSKEY,
  LIFECYCLENUMBER,
  PUBLICVERSION,
  PUBLICOPERATION,
  STATUSCODE,
  VALUEDATE,
  REASONCODE
FROM
  SELECT
    COLLECTIONKEY,
    FLOWSYSKEY,
    LIFECYCLENUMBER,
    PUBLICVERSION,
    PUBLICOPERATION,
    STATUSCODE,
    VALUEDATE,
    REASONCODE,
    COLLECTIONVERSION,
    MAX(COLLECTIONVERSION) OVER (PARTITION BY FLOWSYSKEY) AS MAX_COLLECTIONVERSION
  FROM AA
  WHERE COLLECTIONKEY LIKE '1437023L%'
WHERE COLLECTIONVERSION = MAX_COLLECTIONVERSION;But without knowing a lot more about your data distributions, etc... it's just a guess.