OLAPy stuff crawling a bit

I'm emulating ROLAP in Postgres, populating a sorta-cube from a table with
200 million rows in it. Many aggregates over around 20 fields. Two questions
:

1. Currently, it takes 1000 seconds to build a cube from scratch. What,
other then trivial, tricks can be done to speed up the monstrous SELECT

2. Adding a text field to the source table makes a select really crawl even
when such field isn't in the select. Is this supposed to be so ?