Tuning required for the query

Question

Hi,
I am just wondering what could be the best SQL for the following scenario.
Table 1 Table2
COL1 COL2 COL3 COL1 COL2 COL3
AAA 123 10 AAA 123 12
AAA 123 12
AAA 123 14
AAA 456 10
AAA 456 12
AAA 456 14
Given the above data, I want only the AAA/123/10 and AAA/123/14 records from table1, and leave the others alone.
The tables are huge (200 million) and indexed on all the three columns and my SQL is taking much time to complete.
SQL i have used
select col1, col2, col3 from (
select a.col1, a.col2, a.col3, b.col3 col4 from
table1 a
left join
table2 b
on
a.col1 = b.col1 and
a.col2 = b.col2 and
a.col3 = b.col3
where exists (select 1 from table2 c where a.col1 = c.col1 and a.col2 = c.col2)
) x
where x.col4 is null
Can anyone give me better solution than this?

Answer 1

Depending on your data, there is a good possibility that the intermediate result set, after the outerjoin from table1 to table2, is even more huge than the original tables. This might impact the time it takes to process your query in a negative way.
The next query does something similar, but with a smaller intermediate result sets and is therefore probably a bit faster. At least it's worth the try. Please let us know the impact.
SQL> select t1.col1
  2       , t1.col2
  3       , t1.col3
  4    from table1 t1
  5   where exists
  6         ( select null
  7             from table2 t2
  8            where t2.col1 = t1.col1
  9              and t2.col2 = t1.col2
10         )
11   minus
12  select t2.col1
13       , t2.col2
14       , t2.col3
15    from table2 t2
16  /
COL       COL2       COL3
AAA        123         10
AAA        123         14
2 rows selected.Regards,
Rob.