I have a table with three fields, let's call them username, eventDate and eventType. I have millions of rows of data in it, and I have two indices created for this table already: idx_event which indexes username and eventDate together, and idx_type which only indexes the eventType field. Both indexes support reverse lookup.
I want to run queries that all look like this:
SELECT username, eventDate
FROM myTable
WHERE eventType IN (....)
GROUP BY username, eventDate
The only part that changes from query to query, is the IN part.
I wanted to optimize my query so I ran EXPLAIN for one such query, and here is what I found in my SYSTOOLS.EXPLAIN_OPERATOR table:
If I understand this correctly, it reads from bottom to top; first an index (I think it is the idx_type) is used to fetch the reslut of where part, but then there is a join (I think it is the IN part), and then a SORT and GRPBY which is for the GROUP BY.
Assuming I got it right, it seems that the rate limiting factor is the NLJOIN part. How can I used indices to make it more efficient?
I want to run queries that all look like this:
SELECT username, eventDate
FROM myTable
WHERE eventType IN (....)
GROUP BY username, eventDate
The only part that changes from query to query, is the IN part.
I wanted to optimize my query so I ran EXPLAIN for one such query, and here is what I found in my SYSTOOLS.EXPLAIN_OPERATOR table:
Code:
OPERATOR_ID OPERATOR_TYPE TOTAL_COST IO_COST CPU_COST FIRST_ROW_COST
1 RETURN 635.29 84 1022526.94 635.29
2 GRPBY 635.29 84 1022526.94 635.29
3 TBSCAN 635.29 84 1017026.94 635.29
4 SORT 635.29 84 1010239.94 635.29
5 NLJOIN 635.28 84 981343.19 30.26
6 TBSCAN 0.00 0 720.00 0.00
7 FETCH 30.26 4 92573.96 30.26
8 IXSCAN 22.70 3 80753.96 22.70Assuming I got it right, it seems that the rate limiting factor is the NLJOIN part. How can I used indices to make it more efficient?