On Mon, 28 Apr 2003, Tom Lane wrote:
> Oleg Bartunov <oleg(at)sai(dot)msu(dot)su> writes:
> >> But this is not the whole story because heap pages are normally crammed
> >> full while btree index pages are normally only filled 2/3rds full during
> >> initial creation. (Plus you have to allow for upper b-tree levels, but
> > Are there any benefits from getting btree index pages to be more effective
> > in space usage ? I've read some paper about 98% space usage for Btree.
> Standard theory says that optimal load for a b-tree is 65-70%. We used
> to make CREATE INDEX cram the leaf pages full, but that just resulted in
> a lot of page splits as soon as you did any inserts or updates. And the
> page splits destroy the physical ordering of the index, which negates
> any I/O savings you might have had from fewer pages.
Thanks for comment, I didn't think about that.
> I suppose if you know that the table will be static there might be some
> value in telling CREATE INDEX to pack the index pages full, but I'm not
> sure it's worth the trouble ...
> regards, tom lane
Oleg Bartunov, sci.researcher, hostmaster of AstroNet,
Sternberg Astronomical Institute, Moscow University (Russia)
Internet: oleg(at)sai(dot)msu(dot)su, http://www.sai.msu.su/~megera/
phone: +007(095)939-16-83, +007(095)939-23-83
In response to
pgsql-general by date
|Next:||From: Eric D Nielsen||Date: 2003-04-28 15:23:02|
|Subject: Doxygen/Javadoc-esque SQL DDL documentor?|
|Previous:||From: Ivan||Date: 2003-04-28 15:07:31|
|Subject: encrypt/decrypt problem|