Re: Oracle statistics

> > the data that replaces the old data, is that similar ?
> > I mean: are big tables always big ? small tables always small ? are
> > the indexes always as selective as before ?
> > If that is the case, why don't you store the statistics in a table
> > before dropping your tables and put them back when you recreated the
> > tables ? That saves many hours. By the time yoy notice that the
> > execution plans are wrong, you can always refresh the statistics for
> > the tables that have large deviations from the statistics.
> >
> > Ronald.
> > -----------------------
> > http://ronr.nl/unix-dba
>
> Hi
> the data that gets replaced is not the same, its always something new,
> this procedure was suggested by a Oracle consultant -:) at this place
> they were using a exp/imp before but was very long, then they come up
> with the ideia of ctas which is suposed to be faster then exp/imp, but
> whithn last month I beleive has been giving lots of troubles, this is
> for a client of ours which wants it like this the ctas dropping all
> the tables and then recreating new indexes, tables etc .. which I
> beleive is a waste of time. I also notice that when the ctas is
> running it scans all the tables which takes time as well ...
> I have suggested to do a partition tables, do to same of then been
> very big, but no nes so far ...
> Thanks
> Maria

Maria,
I can understand that the data is not the same. The value of the statistics can be the same despite the fact that the actual data is different. For histograms this can be different but the table and index statistics can still be valid enough for the optimizer to find a good path.

Ronald.