[postgis-users] Is my query planner failing me, or vice versa?

[Markus Schaber]

Hi, Robert,

Robert Burgholzer wrote:

> Does this suggest some type of opportunity to optimize PostgreSQL for
> GIS databases by changing the page size to something larger, say, enough
> to accommodate the 60-80% of your largest geometry column and data row? 

I think that it might help if you _only_ store such large objects.

But usually, you have other tables containing smaller objects and rows,
and for every single row fetch, the whole page has to be loaded.

The same is for index pages. Although you need only a few tens of bytes,
the whole page has to be fetchend in.

So I think that, for all real-world workloads, the additional overhead
for those cases is much larger than the relatively small overhead to
fetch the TOAST pages.

Btw, the best place to discuss this is the Postgresql performance list
at <pgsql-performance at postgresql.org>.

> This may be a dumb question, and it may not be as easy as changing the
> constant and recompiling... I was just wondering.

There are no dumb questions, but only dumb answers :-)

I think (within limits) it is as easy as changing the constant when
recompiling, but I never tried it.

Markus