Rebuild Index VS Drop and Rebuild?

Question

Hey all,
I am currently redesigning a weekly process (weekly coz we pre determined the rate of index fragmentation) for specific indexes that get massive updates. The old process has proved to be able to fix and maintain reports performance.
In this process we rebuild specific indexes using the below command:
Alter index index_name rebuild online;
This command takes around 10 min for selected indexes.
Testing the below took 2 min for 6 or 7 indexes.
Drop Index Index_Name;
Create Index Index_Name on Table_name (Col1, col, ..);
I know that indexes might not be used, and the application performance would be degraded with stale or non-existent stats. But our production and all our test DBs have procedures that daily gather stats on them.
I tested the below script to make sure that execution plan does not change:
SELECT ProductID, ProductName, MfrID FROM PRODUCT WHERE MFRID = 'Mfr1';
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)|
| 0 | SELECT STATEMENT | | 37 | 3737 | 13 (0)|
| 1 | TABLE ACCESS BY INDEX ROWID| PRODUCT | 37 | 3737 | 13 (0)|
| 2 | INDEX RANGE SCAN | PRODUCT_X1 | 37 | | 3 (0)|
dropping PRODUCT_X1 and recreating it only changed the cost to 12.
Gathering the stats again took the cost to 14.
No performance issues were faced and index was still used.
My question is: Is there any oracle recommendation that requires rebuilding the index instead of dropping and recreating it?
Is there any side effect to my approach that I did not consider?
Thank you

Answer 1

Charlov wrote:
I am currently redesigning a weekly process (weekly coz we pre determined the rate of index fragmentation)Nice. Not only have you defined and located index fragmentation but have also measured the rate at which it occurs.
Could you please share your definition of index fragmentation, how you detect it, and how you measure the rate of change of this fragmentation.
I am curious about all this since it can be repeatedly shown that Oracle btree indexes are never fragmented.
http://richardfoote.files.wordpress.com/2007/12/index-internals-rebuilding-the-truth-ii.pdf
The old process has proved to be able to fix and maintain reports performance.Great so you have traces and run time statistics from before and after the rebuild that highlight this mysterious fragmentation and show how the fragmentation caused the report to be slow, details what effects the rebuild had that caused the reports to perform better.
Please share them as these would be an interesting discussion point since no one has been able to show previously how an index rebuild caused a report to run faster or even show the fragmentation that caused it to be slow in the first place.
I mean it would be a pity if the report was just slow because of an inefficient plan and compressing an index or two that probably shouldn't be used in teh first place appears to temporarily speed it up. Could you imagine rebuilding indexes every week, because some developer put the wrong hint in a query? That would be pretty funny.