On 12 November 2013 21:41, Nicolas Barbier <nicolas.barbier@gmail.com> wrote:
> Look-up speed is as follows: Each look-up must look through all
> B-trees.
That can be optimised by using a min max approach, so we need only
look at sub-trees that may contain data.
> Index size: I think (didn’t calculate) that the combined size of the
> B-trees will be about the same as (a little bit more than) the size of
> a single big B-tree containing the same entries.
Agreed
> Major missing piece in PostgreSQL (I think):
>
> * Functionality to merge K indexes into one (bigger) combined index.
Good analysis.
I would add that it is possible to optimise large DELETEs from a table
if complete sub-trees of the btree can be easily removed. This for me
would be the compelling benefit of this approach.
I still think we need to look at this from a query perspective though.
We need to check whether there is a class of real world queries that
are not well optimised by minmax indexes, or cannot be made to be in
future releases. For example, large DELETEs from a table are almost
trivially optimised for min max.
-- Simon Riggs http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
