Re: Cost of Access Plan using Bitmap Indexes

Hi,

 just to explain this last paragraph...
in the thread "Help me read my 10053 trace file " on this newsgroup Jonathan said:

<snip>
If the selectivity from the index to the table is 0.5, then I would expect Oracle to work out that the table block hits will be AT LEAST 0.5 * number of blocks in table (viz 7839/2), but the figure is usually higher, based on the clustering factor, rather than blocks in table.

However, the clustering_factor for bitmap indexes seems to have nothing to do with the clustering factor for b-tree indexes, I've found recently that it seems to match the number of distinct bitmap values. Nevertheless, there must be some data lurking somewhere which has made your total cost

    134 + 5089
rather than

    134 + 3920
But in either case, this is larger than the 1190 cost of tablescan.
<snip>

what I was expecting in my case was
number of block * selectivity calculated 209747*0.09=18877.23
more or less
is it possible that I am miscalculating my index selectivity? idx1=cost1 and (idx2=cost2 or idx2=cost3) 0.09=(FF1=0.25)*(FF2=0.2+FF2-FF2*FF2)

is this ok am i doing something stupid here?

Thanks again
Fabrizio

>
> from another post of you that I found I would expected a value of
> 18000 more or less... but twice that???
>
> any way if anyone can give me an idea I would really much appreciate!!!
>
> Thanks again
> Fabrizio

-- 
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG