SQL query taking time

After creating the index the cost was reduced but the CPU % got increased :

create index  MGPO_ORG_HIERARCHY_NAME_IDX on MS_GRC_PARENT_ORGS(ORG_HIERARCHY_NAME);
create index  MFDO__OEI_USER_NAME_IDX on MS_FLOW_DOWN_ORG(ORG_ENTITY_ID,USER_NAME );

---OUTPUT 
index MGPO_ORG_HIERARCHY_NAME_IDX created.
index MFDO__OEI_USER_NAME_IDX created.

explain plan for
 SELECT org_entity_name
       ,org_entity_id
       ,org_hierarchy_name
       ,division
       ,area
       ,organization
   FROM ms_grc_parent_orgs mgpo
  WHERE EXISTS (SELECT 1
           FROM ms_flow_down_org
          WHERE org_entity_id = mgpo.org_entity_id
                AND user_name = 'E131094')
        AND org_hierarchy_name = 'Level 4 - Department';

select * from table(dbms_xplan.display);

Plan hash value: 1039081615
 
-------------------------------------------------------------------------------------------------
| Id  | Operation             | Name                    | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |                         |     1 |   141 |   229   (2)| 00:00:03 |
|*  1 |  HASH JOIN SEMI       |                         |     1 |   141 |   229   (2)| 00:00:03 |
|*  2 |   TABLE ACCESS FULL   | MS_GRC_PARENT_ORGS      |   161 | 20930 |     5   (0)| 00:00:01 |
|*  3 |   INDEX FAST FULL SCAN| MFDO__OEI_USER_NAME_IDX |  3524 | 38764 |   223   (1)| 00:00:03 |
-------------------------------------------------------------------------------------------------
 
Predicate Information (identified by operation id):
---------------------------------------------------
 
   1 - access("ORG_ENTITY_ID"="MGPO"."ORG_ENTITY_ID")
   2 - filter("ORG_HIERARCHY_NAME"='Level 4 - Department')
   3 - filter("USER_NAME"='E131094')
 
Note
-----
   - 'PLAN_TABLE' is old version

ALTER INDEX MGPO_ORG_HIERARCHY_NAME_IDX
  REBUILD COMPUTE STATISTICS;

ALTER INDEX MFDO__OEI_USER_NAME_IDX
  REBUILD COMPUTE STATISTICS;	

Still the plan was same: 

Plan hash value: 1039081615
 
-------------------------------------------------------------------------------------------------
| Id  | Operation             | Name                    | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |                         |     1 |   141 |   229   (2)| 00:00:03 |
|*  1 |  HASH JOIN SEMI       |                         |     1 |   141 |   229   (2)| 00:00:03 |
|*  2 |   TABLE ACCESS FULL   | MS_GRC_PARENT_ORGS      |   161 | 20930 |     5   (0)| 00:00:01 |
|*  3 |   INDEX FAST FULL SCAN| MFDO__OEI_USER_NAME_IDX |  3524 | 38764 |   223   (1)| 00:00:03 |
-------------------------------------------------------------------------------------------------
 
Predicate Information (identified by operation id):
---------------------------------------------------
 
   1 - access("ORG_ENTITY_ID"="MGPO"."ORG_ENTITY_ID")
   2 - filter("ORG_HIERARCHY_NAME"='Level 4 - Department')
   3 - filter("USER_NAME"='E131094')
 
Note
-----
   - 'PLAN_TABLE' is old version

Select allOpen in new window

and the time of the actual query?

is taking around 9.5-10.5 Secs

The FIRST_ROWS hint has nothing to do with the number of rows to be returned.  Essentially the default behavior of the CBO is what is the ALL_ROWS hint.  This tends toward full table scans.  A FIRST_ROWS hint tends toward nested loops.  The emphasis is on getting back the first rows the fastest.

Given the size of the tables, it doesn't look to me like you would ever get to use an index on the outer table.  What if you switched the order of the 2 columns in the MFDO__OEI_USER_NAME_IDX index?  That might get you an index range scan rather than a full scan.  That might save some time.

The FIRST_ROWS hint has nothing to do with the number of rows to be returned.

That is not fully true, as shown here:

https://docs.oracle.com/cd/E11882_01/server.112/e41084/sql_elements006.htm#BABGCCFG

The FIRST_ROWS hint instructs Oracle to optimize an individual SQL statement for fast response, choosing the plan that returns the first n rows most efficiently. For integer, specify the number of rows to return.

So, again: Obviously this "integer" parameter is not optional. What does Oracle / the CBO do if you ommit this value?!

>>I thought this hint needs a certain number of rows like: ?!?

Never knew about the number parameter.  I try to avoid hints and let the optimizer do its magic.  The number is optional.  Not sure what the benefit to specifying the number is.

You would just need to experiment to see any possible benefits.

Never knew about the number parameter.  I try to avoid hints and let the optimizer do its magic.

I totally agree with you ;-)
But, nevertheless, if this parameter is optional then the official docs are wrong about this one...

I never used the integer hint.  It is a optimizer path hint.  It tells the optimizer which join method to try to use.  ALL_ROWS will tend to use hash joins and FIRST_ROWS will tend to use nested loops.

Got it:
https://docs.oracle.com/cd/B19306_01/server.102/b14200/sql_elements006.htm

The FIRST_ROWS hint specified without an argument, which optimizes for the best plan to return the first single row, is retained for backward compatibility and plan stability only.

Obviously deprecated...

We tend to not use hints.  The optimizer has gotten much better so we don't find that we need to use them that often.

FIRST_ROWS is a blanket hint, we tried not to use it.  Hinting the optimizer to use specific indexes was a much better way to go than using this hint.

Getting way off topic.

I think the more important part of my post was:  AND rownum=1

>>> I have the below SQL which is taking 12 sec to execute when attached to a web application:


What is the execution time of the query by itself - without the web app?

after making the changes as suggested :

 SELECT org_entity_name
       ,org_entity_id
       ,org_hierarchy_name
       ,division
       ,area
       ,organization
   FROM ms_grc_parent_orgs mgpo
  WHERE EXISTS (SELECT /*+ OPT_PARAM('optimizer_dynamic_sampling' 6) */ 1
           FROM ms_flow_down_org
          WHERE org_entity_id = mgpo.org_entity_id
                AND user_name = 'E131094'
                and rownum =1)
        AND org_hierarchy_name = 'Level 4 - Department';

Select allOpen in new window

Now it is taking 0.7 sec

you should not need rownum=1 that's just a wasted step in the plan

the HASH JOIN SEMI was already doing the same thing (or should have anyway - that's what the "SEMI" means)

after your remove the rownum, get full plan with the predicate clauses

This question was really a learning for me. Thanks a lot experts for all your comments.


Thanks,
Sloba

Awesome inputs provided on this question...