Explain plan : need advice in improving performance

Question

Hi,
DB : 10.2.0.4
platform: Solaris
SGA: 8 G
one of my query is taking too much time, explain plan gives below output.
kindly advise what i can do to improve the performance
PLAN_TABLE_OUTPUT
Plan hash value: 430877948
| Id  | Operation                    | Name                          | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |                               |  8143 | 1081K| 31130   (1)| 00:06:14 |
|*  1 | FILTER | |       |       |            |          |
|*  2 | TABLE ACCESS BY INDEX ROWID| ID_TICKET_DETAILS             |  8143 | 1081K|   494   (0)| 00:00:06 |
|*  3 | INDEX SKIP SCAN           | TKT_IDX_21                    |     1 | |   493   (0)| 00:00:06 |
|*  4 | TABLE ACCESS BY INDEX ROWID| ID_DELIVERY_DEBIT_SLIP_DETAIL |     1 | 34 |     3   (0)| 00:00:01 |
|   5 | NESTED LOOPS              |                               |     2 | 124 |     7   (0)| 00:00:01 |
|*  6 | TABLE ACCESS FULL        | ID_DELIVERY_DEBIT_SLIP_HEADER | 32243 | 881K|     2   (0)| 00:00:01 |
|*  7 | INDEX RANGE SCAN         | DSD_DELIVERY_DEBIT_UKEY       |     1 | |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter( NOT EXISTS (SELECT /*+ */ 0 FROM "ID_DELIVERY_DEBIT_SLIP_HEADER"
"ID_DELIVERY_DEBIT_SLIP_HEADER","ID_DELIVERY_DEBIT_SLIP_DETAIL" "ID_DELIVERY_DEBIT_SLIP_DETAIL" WHERE
"DSH_DOCUMENT_NUMBER"="DSD_DOCUMENT_NUMBER" AND "DSH_DOCUMENT_TYPE"="DSD_DOCUMENT_TYPE" AND
"DSH_COMPANY"="DSD_COMPANY" AND "DSD_TICKET_NUMBER" IS NOT NULL AND
LNNVL(:B1||:B2<>"DSD_AIRLINE"||"DSD_TICKET_NUMBER") AND "DSH_DELIVERY_DEBIT"='DEBIT'))
   2 - filter((:1 IS NULL OR "TICKET_AIRLINE"=:2) AND "TICKET_REFERENCE_2" IS NULL AND
"TICKET_RECEIPT_NUMBER" IS NULL AND "TICKET_CARD_RECEIPT_NUMBER" IS NULL AND
"TICKET_SYSTEM_DOC_NUMBER" IS NULL)
   3 - access("TICKET_REFERENCE_1" IS NULL)
       filter("TICKET_REFERENCE_1" IS NULL AND TO_NUMBER("TICKET_COMPANY")=1)
   4 - filter("DSD_TICKET_NUMBER" IS NOT NULL AND LNNVL(:B1||:B2<>"DSD_AIRLINE"||"DSD_TICKET_NUMBER"))
   6 - filter("DSH_DELIVERY_DEBIT"='DEBIT')
   7 - access("DSH_COMPANY"="DSD_COMPANY" AND "DSH_DOCUMENT_TYPE"="DSD_DOCUMENT_TYPE" AND
"DSH_DOCUMENT_NUMBER"="DSD_DOCUMENT_NUMBER")
Note
   - SQL profile "SYS_SQLPROF_014f902e2ea4c002" used for this statement

Answer 1

some comments:
it would be more simple to read the plan with indentatitions: you could use a fixed-width font
it's hard to tell much about the plan without seeing the corresponding query (though in this case the predicate section gives some information on the query - especially step 1)
the plan shows the use of a sql profile: so the CBO uses additional statistics to generate the plan
in step 3 there is an index skip scan: that's only a good idea if there are few distinct values for the leading column of the index and the selectivity of "TICKET_REFERENCE_1" IS NULL is good
in step 6 there is a Full Table Scan for the driving table of a nested loops join: the cost value for the scan is very small and so is the cost for the complete NL join - and that could be misleading
I would use the gather_plan_statistics hint to get a plan with rowsource statistics to check if the cardinalities the CBO works with are correct. If they are not you could try to disable the profile (or create a new profile; of course after checking who created the profile and for what reasons). With an accurate sql profile the CBO should have enough information to create an accurate plan in most cases.
Regards
Martin