Speed up data retrieval

Question

I have around millions of data in oracle inventory which i need to show in the website. currently we r by running a procedure that get this information and display the information to user. but this take lots of time. is there any other method which would reduce the time take to fetch the data.

Answer 1

Nazoor
First, is a 10 minute runtime for your query really a problem? How many times/day do you run it? It may be better to wait 10 mins here than to spend time and effort to get it down to a few seconds - at the same time as making your critical load slower.
What is the range of values for the number you are selecting on? is it random, or is it something like a status value? Is it skewed?
How many rows are you loading per day? Is each partition a single day, or several days. How much data do you have overall (ie what range of dates).
A possible solution might be:
1) split your data in two
- latest data (ie today) - lets call that A
- historic data (everything else - your existing partitioned table - lets call B (with partitions B1, B2, ... )
2) Index B by your number column (lets call it X). DON'T index A.X
3) create a UNION ALL view over A and B which you can use for queries that need to include today's latest data
4) When you query by the number, the CBO should divide into two queries
- on A, which can do a (small) tablescan)
- on B which can use the index(es)
5) Loading into A is unaffected (no additional index)
6) At end of each day, create an index on A.X (creating it once is quicker than loading with an index many times) and then:
7) exchange partition A into B, and create a new A for tomorrow. (or simply copy A into B using INSERT INTO / SELECT FROM - that would make sure that any "late" data in A got sorted into the appropriate partition in B - but does mean you are double-handling that data).
Would something like that work for you?
Regards Nigel