I've also now done a test utilising the script supplied by tip of the week
and the script from metalink. it makes interesting reading.
Tip of the week - taking a monthly time window i.e. I'm assuming that I have
a monthly maintenance window - suggests that I should rebuild 197 indexes in
this database. The metalink script suggests I rebuild 2.
--
Niall Litchfield
Oracle DBA
Audit Commission UK
"Niall Litchfield" <n-litchfield_at_audit-commission.gov.uk> wrote in message
news:3f94fbb9$0$246$ed9e5944_at_reading.news.pipex.net...
> "Chris Boyle" <cboyle_at_hargray.com> wrote in message
> news:VP%kb.31047$N94.16173_at_lakeread02...
> > OK, I admit it. I do not understand the problem. Would someone please
> > explain why this deserves derision?
>
> The code attempts to build a list of indexes to be rebuilt based on 2
> criteria
>
> 1. How old the index is.
> 2. How big the index is
>
> so that you rebuilt your oldest indexes first, but allow for the fact that
> you may not have enough time in the maintenance window to rebuild every
> index.
>
> The underlying assumption is that rebuilding your indexes is a Good Thing.
> In general it is just a waste of time and effort. Even the advice from
> Oracle suggests that you should only rebuild when you have a depth of 4 or
> greater and more than 20% of the index entries are deleted entries. Note
> 232819.1 on metalink.
>
> So, don't rebuild your indexes unless you really need to and decide when
you
> need to based on sensible criteria not time since rebuild. I'm pretty sure
> that Howard and Richard have some suggestions as to whether the criteria
> suggested by Oracle above are sensible or not.
>
>
> --
> Niall Litchfield
> Oracle DBA
> Audit Commission UK
>
>
Received on Tue Oct 21 2003 - 04:58:16 CDT
