Comments on: Access Path Scientific Analysis Part III http://www.remotedbaexperts.com/Blog/2009/12/access-path-scientific-analysis-part-iii/ Remote DBA Experts Blog Fri, 23 Sep 2011 17:39:48 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Chris Foot http://www.remotedbaexperts.com/Blog/2009/12/access-path-scientific-analysis-part-iii/comment-page-1/#comment-123 Chris Foot Thu, 03 Dec 2009 15:34:16 +0000 http://www.remotedbaexperts.com/Blog/?p=794#comment-123 You are exactly correct Peter. I need to add that to the post! That is a definite problem with index monitoring, stats will flip the "used switch" You are exactly correct Peter. I need to add that to the post! That is a definite problem with index monitoring, stats will flip the “used switch”

]]> By: Narendra http://www.remotedbaexperts.com/Blog/2009/12/access-path-scientific-analysis-part-iii/comment-page-1/#comment-122 Narendra Thu, 03 Dec 2009 10:35:12 +0000 http://www.remotedbaexperts.com/Blog/?p=794#comment-122 Chris, Thanks for yet another nice post. I hope you don't get distracted too much by your non-technical activities. :) Chris,

Thanks for yet another nice post. I hope you don’t get distracted too much by your non-technical activities. :)

]]> By: Peter Moore http://www.remotedbaexperts.com/Blog/2009/12/access-path-scientific-analysis-part-iii/comment-page-1/#comment-120 Peter Moore Wed, 02 Dec 2009 22:54:27 +0000 http://www.remotedbaexperts.com/Blog/?p=794#comment-120 I've always found that index monitoring is an idea that sounds a lot more useful than it actually is. In order to find out if an index is ever really used, you have to monitor it for a long period of time - just in case it's one of those indexes that's only used in weekly/monthly/annual reports or batch runs. However, if you analyze a table (or run any of the gather_stats routines) then any indexes attached to the table will be marked as 'used', even though they haven't actually been used in a real sense. Which immediately invalidates the monitoring which was so carefully setup. I’ve always found that index monitoring is an idea that sounds a lot more useful than it actually is.

In order to find out if an index is ever really used, you have to monitor it for a long period of time – just in case it’s one of those indexes that’s only used in weekly/monthly/annual reports or batch runs.

However, if you analyze a table (or run any of the gather_stats routines) then any indexes attached to the table will be marked as ‘used’, even though they haven’t actually been used in a real sense. Which immediately invalidates the monitoring which was so carefully setup.

]]>