Re: Transactions requires much more UNDO space than expected....

Sybrand Bakker wrote:
> On 19 Jun 2006 10:43:24 -0700, bdbafh_at_gmail.com wrote:
>
> >> 2 Observing the redo per record is still unreliable (because of the
> >> rebalancing of indexes), but it is the most reliable method. So indeed
> >> you need to multiply the redo per record by the number of records.
> >> Would be nice to know whether you are less off from the actual redo by
> >> this calculation.
> >
> >Please explain what you mean by "rebalancing of indexes".
> >
>
> >> 3 If you populate an empty index you force Oracle to rebalance the
> >> B*-tree often.
> >
> >Please explain whay you mean by "rebalance the B*-tree".
>
> Should I explain what a balanced tree is, and what bucket split is?
> Is that a serious question, or are you trying to have me abstract
> common information textbooks?
>
> --
> Sybrand Bakker, Senior Oracle DBA

To me, if something is "re-balanced" that would imply that it was balenced and then out of balance. I am not looking for you to extract common information, just to discuss how you think that the index was "out of balance" so as to need "re-balancing" or "balancing". (50-50 split and 90-10 splits aside).

-bdbafh Received on Mon Jun 19 2006 - 14:56:23 CDT