How many columns to index

Question

Hi,
I have a table t1 which has nearly 20000 rows. It is accessed by a query which has 3 columns in its where clause lets say col1, col2 and col3. the table doesn't have any index so it does a full table scan. Now if I put an index on all the 3 columns it uses the index and avoids the FTS. Also if I index just 2 columns instead of three then also it uses index (something that I don't know why). my question is should I index the two coulmns or three columns? the third column which i left out has nearly 8000 distinct values.
the query is of the form:
select col5 from table1 where col1=value1 and col2=value2 and col3=value3 and rownum=1
the execution plan with index on three columns is:
Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
0 | SELECT STATEMENT | | 1 | 24 | 2 (0)| 00:00:01 |
* 1 | COUNT STOPKEY | | | | | |
* 2 | TABLE ACCESS BY INDEX ROWID| table1     | 1 | 24 | 2 (0)| 00:00:01 |
* 3 | INDEX RANGE SCAN | ind-3col | 1 | | 1 (0)| 00:00:01 |
Predicate Information (identified by operation id):
2 - access("col1"=6003 AND "col2"=1532 AND
"col3"=267)
the execution plan with index on two columns is:
======
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
| 0 | SELECT STATEMENT | | 1 | 24 | 2 (0)| 00:00:01 |
|* 1 | COUNT STOPKEY | | | | | |
|* 2 | TABLE ACCESS BY INDEX ROWID| table1 | 1 |     24 | 2 (0)| 00:00:01 |
|* 3 | INDEX RANGE SCAN | ind-2col | 1 | | 1 (0)| 00:00:01 |
Predicate Information (identified by operation id):
1 - filter(ROWNUM=1)
2 - filter("col3"=267)
3 - access("col1"=6003 AND "col2"=1532)
I don't know if I should index on 2 columns or 3 columns...can someone suggest?
Thanks
Edited by: orausern on Feb 8, 2010 5:28 AM
Edited by: orausern on Feb 8, 2010 5:29 AM

Answer 1

Hi,
To index or not to index on columns would totally depends on type of query(update/delete/select) being run on that table.
So you need to check all/majority of the query that would be running on this table to decide which column should be indexed.
eg:
If i have just one query which you posted that would be running against this table then I would have index on all the three columns.
But only 10-20% of the query would be making use of all the three columns, and rest 70-80% of the queries are making use of same 2 out of these 3 columns then I would be indexing only 2 columns.
I suppose you would also like to understand that queries also determines which column would be leading column in index.
Index makes select queries to run faster in most of the case, but it also makes insert statements run slower and utilized space. So keeping all this factor in mind it should be decided on which columns and in which order should we index.
Regards
Anurag