Re: Performance problem with deleteing large number of rows from a table

"Richard Foote" <richard.foote_at_bigpond.nospam.com> wrote in message news:QTlDg.11004$rP1.1210_at_news-server.bigpond.net.au...
>> 1. If the table is index organized am I right to assume that this will
>> dramatically change the calculation? Using your example we would get rid
>> of 1,000,000 LIO to table blocks. In that case the index access seems to
>> be always superior.
>
> Well yes and no. Yes you no longer have to visit a heap table structure as
> such (assuming you have no overflow area defined), however assuming the
> table data and volume is the same, you'll still have the 1,000,000 blocks.
> They'll just be part of the leaf index structure. And you still need the
> various layers of branch blocks sitting above the leaf blocks so a FTS (or
> a full index scan as is kinda the case of a IOT) will actually be more
> expensive as a result of having to also access these additional branch
> blocks during this process.
>

I meant to say a *fast* full index scan.

Cheers

Richard Received on Sun Aug 13 2006 - 08:50:33 CDT