Select query performance is very slow

Question

Could you please explain me about BITMAP CONVERSION FROM ROWIDS
Why the below query going for two times BITMAP CONVERSION TO ROWIDS on the same table.
SQL> SELECT AGG.AGGREGATE_SENTENCE_ID ,
  2         AGG.ENTITY_ID,
  3         CAR.REQUEST_ID REQUEST_ID
  4    FROM epic.eh_aggregate_sentence agg ,om_cpps_active_requests car
  5   WHERE car.aggregate_sentence_id =agg.aggregate_sentence_id
  6  AND car.service_unit = '0ITNMK0020NZD0BE'
  7  AND car.request_type = 'CMNTY WORK'
  8  AND agg.hours_remaining > 0         
  9  AND NOT EXISTS (SELECT 'X'
10                    FROM epic.eh_agg_sent_termination aggSentTerm
11                   WHERE aggSentTerm.aggregate_sentence_id = agg.aggregate_sentence_id
12                     AND aggSentTerm.date_terminated <= epic.epicdatenow);
Execution Plan
Plan hash value: 1009556971
| Id  | Operation                           | Name                       | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT                    |                            |     5 |   660 |    99   (2)| 00:00:02 |
|*  1 |  HASH JOIN ANTI                     |                            |     5 |   660 |    99   (2)| 00:00:02 |
|   2 |   NESTED LOOPS                      |                            |       |       |            |          |
|   3 |    NESTED LOOPS                     |                            |     7 |   658 |    95   (0)| 00:00:02 |
|*  4 |     TABLE ACCESS BY INDEX ROWID     | OM_CPPS_ACTIVE_REQUESTS    |    45 |  2565 |    50   (0)| 00:00:01 |
|   5 |      BITMAP CONVERSION TO ROWIDS    |                            |       |       |            |          |
|   6 |       BITMAP AND                    |                            |       |       |            |          |
|   7 |        BITMAP CONVERSION FROM ROWIDS|                            |       |       |            |          |
|*  8 |         INDEX RANGE SCAN            | OM_CA_REQUEST_REQUEST_TYPE |   641 |       |    12   (0)| 00:00:01 |
|   9 |        BITMAP CONVERSION FROM ROWIDS|                            |       |       |            |          |
|* 10 |         INDEX RANGE SCAN            | OM_CA_REQUEST_SERVICE_UNIT |   641 |       |    20   (0)| 00:00:01 |
|* 11 |     INDEX UNIQUE SCAN               | PK_EH_AGGREGATE_SENTENCE   |     1 |       |     0   (0)| 00:00:01 |
|* 12 |    TABLE ACCESS BY INDEX ROWID      | EH_AGGREGATE_SENTENCE      |     1 |    37 |     1   (0)| 00:00:01 |
|  13 |   TABLE ACCESS BY INDEX ROWID       | EH_AGG_SENT_TERMINATION    |    25 |   950 |     3   (0)| 00:00:01 |
|* 14 |    INDEX RANGE SCAN                 | DATE_TERMINATED_0520       |     4 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - access("AGGSENTTERM"."AGGREGATE_SENTENCE_ID"="AGG"."AGGREGATE_SENTENCE_ID")
   4 - filter("CAR"."AGGREGATE_SENTENCE_ID" IS NOT NULL)
   8 - access("CAR"."REQUEST_TYPE"='CMNTY WORK')
  10 - access("CAR"."SERVICE_UNIT"='0ITNMK0020NZD0BE')
  11 - access("CAR"."AGGREGATE_SENTENCE_ID"="AGG"."AGGREGATE_SENTENCE_ID")
  12 - filter("AGG"."HOURS_REMAINING">0)
  14 - access("AGGSENTTERM"."DATE_TERMINATED"<="EPIC"."EPICDATENOW"())now this query is giving the correct result, but performance is slow.
Please help to improve the performance.
SQL> desc epic.eh_aggregate_sentence
Name                                      Null?    Type
ENTITY_ID                                          CHAR(16)
AGGREGATE_SENTENCE_ID                     NOT NULL CHAR(16)
HOURS_REMAINING                                    NUMBER(9,2)
SQL> desc om_cpps_active_requests
Name                                      Null?    Type
REQUEST_ID                                NOT NULL VARCHAR2(16)
AGGREGATE_SENTENCE_ID                              VARCHAR2(16)
REQUEST_TYPE                              NOT NULL VARCHAR2(20)
SERVICE_UNIT                                       VARCHAR2(16)
SQL> desc epic.eh_agg_sent_termination
Name                                      Null?    Type
TERMINATION_ID                            NOT NULL CHAR(16)
AGGREGATE_SENTENCE_ID                     NOT NULL CHAR(16)
DATE_TERMINATED                           NOT NULL CHAR(20)
.

Answer 1

user10594152 wrote:
Thanks for your reply.
Still i am getting same problemIt is not a problem. Bitmap conversion usually is a very good thing. Useing this feature the database can use one or several unselective b*indexes. Combine them and do a kind of bitmap selection. THis should be slightly faster than a FTS and much faster than a normal index access.
Your problem is that your filter criteria seem to be not very usefull. Whcih is the criteria that does the best reduction of rows?
Also any kind of NOT EXISTS is potentiall not very fast (NOT IN is worse). You can rewrite your query with an OUTER JOIN. Sometimes this will help, but not always.
SELECT AGG.AGGREGATE_SENTENCE_ID ,
        AGG.ENTITY_ID,
        CAR.REQUEST_ID REQUEST_ID
FROM epic.eh_aggregate_sentence agg
JOIN om_cpps_active_requests car ON ar.aggregate_sentence_id =agg.aggregate_sentence_id
LEFT JOIN epic.eh_agg_sent_termination aggSentTerm ON aggSentTerm.aggregate_sentence_id = agg.aggregate_sentence_id and aggSentTerm.date_terminated <= epic.epicdatenow
WHERE car.service_unit = '0ITNMK0020NZD0BE'
AND car.request_type = 'CMNTY WORK'
AND agg.hours_remaining > 0         
AND aggSentTerm.aggregate_sentence_id is nullEdited by: Sven W. on Aug 31, 2010 4:01 PM