On [Thu, 27.08.2009 14:05], Dmitry Yemanov wrote:

> Daniel Vogelbacher wrote:
> >
> > You mean that the first query cleans up the indicies
> > and after gc has finished, the query runs faster?
>
> I mean that after deletion and commit, select count(*) garbage collects
> not only data pages but also index pages. So existence of indices does
> affect this query, while you were suggesting the opposite.

In one of your last mails you wrote that sweep invokes GC.
If I run a manual sweep every night, the select query
should run fast, isn't it?



Daniel

--
web: http://daniel.vogelbacher.name
irc: cytrinox @ (freenode|ircnet|quakenet)