Help in Tuning this Query

Question

Hi,
I have a query in my proj where the same table is looked up twice in the same query.
Can anybody suggest in improving the performance of this query?
select * from table1 a1 where (a1.column1, a1.column2, a1.column3, a1.column4, a1.column5, a1.column6, a1.column7, a1.column8, a1.column9,
a1.column10) in ( select a2.column1, a2.column2, a2.column3, a2.column4, a2.column5, a2.column6, a2.column7, a2.column8, a2.column9,
a2.column10 from table1 a2 where column20 = '<condn>')
The table1 used here is same in outer query as well as the sub query. this is a example of what we use here, and the table1 contains 30 million rows. Though, creating index with 10 columns can be a option, we already have a unique index with 11 columns(which includes 10 from this query) and will that be helpful in anyway? or the same existing index can be forced?
Thanks a lot for ur time

Answer 1

Depending on the selectivity of column20 I am not sure Index is the best way to go. It might perform better with two full scans and a hash-join.
Anyway, I do prefer the syntax:
select /*+ leading(a2) */ a1.*
from table1 a1, table1 a2
where a2.column20 = '<condn>'
and a1.column1 = a2.column1
and a1.column2 = a2.column2
and  a1.column3 = a2.column3
and  a1.column4 = a2.column4
and  a1.column5 = a2.column5
and  a1.column6 = a2.column6
and  a1.column7 = a2.column7
and  a1.column8 = a2.column8
and  a1.column9 = a2.column9
and  a1.column10 = a2.column10;I've added a leading hint to tell oracle that the start table is a2. Might be useless.
Ensure your stats are up to date. You might need histograms here if your column20 is skewed.
Hope this helps,
François
Edited by: Francois Berger on Oct 24, 2008 1:47 AM