On Wed, Oct 29, 2008 at 9:18 PM, Kevin Galligan <kgalligan@gmail.com> wrote:
> I'm approaching the end of my rope here. I have a large database.
> 250 million rows (ish). Each row has potentially about 500 pieces of
> data, although most of the columns are sparsely populated.
>
*snip*
>
> So, went the other direction completely. I rebuilt the database with
> a much larger main table. Any values with 5% or greater filled in
> rows were added to this table. Maybe 130 columns. Indexes applied to
> most of these. Some limited testing with a smaller table seemed to
> indicate that queries on a single table without a join would work much
> faster.
>
> So, built that huge table. now query time is terrible. Maybe a
> minute or more for simple queries.
Are indexes on sparsely populated columns already handled efficiently,
or could partial indexes with only non-null values improve things?
Isak