[xquery-talk] performance figures for xquery processors
yuin at bb.din.or.jp
Wed Feb 8 19:55:03 PST 2006
"Michael Kay" <mhk at mhk.me.uk> wrote:
> > In my laptop, Saxon can't load 100M xml document.
> > it is because memory space (total 768M/free about 250M) is
> > insufficient, but it is scaling topic. That was why I put
> > the issue.
> How much memory did you allocate to the Java VM? Usually -Xmx512M -Xms512M
> should be sufficient for a 100Mb document, though if the processing is
> non-trivial it would be better to allocate more.
tested with -Xmx256m, -Xmx384m, "without -Xmx"..
I don't remember whether tested with -Xmx512m, sorry.
# Of course, I know with more memory spaces it'll run.
> > Object creation is still heavy even in latest JVM
> > and each Object has at least 8 bytes! memory footprint,
> > this costs.
> This is why Saxon doesn't represent nodes of the tree using objects. The
> memory used by the tinytree is about 20bytes per node plus 2 bytes per
> character of text and attribute content. Attribute-intensive documents may
> take more space because attribute values are held as String objects.
If so, array expansion (re-allocation) might have been cost
in the last test, while I have no proof.
# Though it is from my experience..
Nara Institute of Science and Technology (NAIST)
Makoto YUI <yuin405 AT gmail.com>
More information about the talk