Re: Self-generating statistics?

Mario Weilguni <mweilguni(at)sime(dot)com> writes:
> Many of the performance problems on pgsql-general seem to be related
> to the fact that no analyze is performed after the creation on the
> tables

Well, there are lots of other ways an incompetent DBA can screw up a
database. The need to VACUUM and ANALYZE is stated clearly in the
docs. Providing workarounds for negligence isn't the right path to get
started down, IMHO.

That said, the general idea of a self-tuning database system has
merit, IMHO. For example, this paper proposes a histogram data
structure that can be updated fairly cheaply based on data gathered
from query execution:

http://citeseer.nj.nec.com/255752.html

A bunch of industry players (IBM, Microsoft, etc.) are putting some
work into this area (IBM calls it "autonomic computing", for
example). It might be an interesting area to look at in the future...

Cheers,

Neil

--
Neil Conway <neilc(at)samurai(dot)com> || PGP Key ID: DB3C29FC