On Mon, 2008-11-10 at 11:08 -0500, Tres Seaver wrote:
> > Index Name |Type |Avg Time |Calls/second
> > ==============================================================
> > object_implements |KeywordIndex |0.2172234| 4.6
> This is clearly not the same issue as the other KeywordIndexes: in
> fact, I am astonished that anybody would be using a KeywordIndex for
> this at all. I would suspect that the real problem here is in the
> appliation, rather than the index itself.
> > getEffective_or_creat|DateIndex |0.1941770| 5.15
> > effectiveRange |DateRangeIndex |0.0086295| 115.88
> > allowedRolesAndUsers |KeywordIndex |0.0069754| 143.36
> Hmm, I'm surprised there: what query is being passed to 'apply_index'
> for this call?
Well it is not really performing badly at 6ms?
> > path |ExtendedPathIndex|0.0040614| 246.22
> I don't trust the EPI implementation at all.
> > portal_type |FieldIndex |0.0025984| 384.84
> This one is surprising: its performance should be pretty similar to
> other FieldIndexes (e.g., 'review_state') which map a controlled
> vocabulary onto the entire corpus. Was the query different than
> 'review_state' (e.g., multi-valued vs. single-valued)?
It's still not bad at 2ms. It has a lot more keys than review_state
> > SearchableText |ZCTextIndex |0.0007645| 1308.04
> > sourceUID |FieldIndex |0.0004886| 2046.31
> Probably bogus, but I don't know how it is used.
I'm not really worried about indexes beyond this point - they're all
returning results in less than a millisecond.
> Can you provide information on the corpus / configuration / test plan
> you used to generate these results?
It's basically a Plone site with 300,000 remember based users and
roughly 150,000 documents and images indexed.
Upfront Systems http://www.upfrontsystems.co.za
Zope-Dev maillist - Zope-Dev@zope.org
** No cross posts or HTML encoding! **
(Related lists -