From: Wouter H. <ws...@gm...> - 2011-02-11 13:17:55
|
Thanks Wolfgang Splitting collections should be no problem. About querying: so I can just query the top collection using a wildcard? I read somewhere that querying using specific fields with // is faster, but this was also a 1.4 comment. I see the 1.4.1 changelog is up again. I will have a look. 2011/2/11 Wolfgang Meier <wol...@ex...> > You probably have too many documents in a single collection. In 1.4, the > time to update a single document grows proportionally with the size of the > collection. This has something to do with the way indexes are organized. > There are several possible solutions: > > 1) split your main collection into several sub-collections. This has no > effect on queries since exist will always automatically include all > sub-collections in a query to the main collection. > > 2) remove the entire collection before updating it. > > 3) use trunk, which addresses those problems and should be many times > faster when updating large collections. I would actually like to port those > changes back to 1.4 once I'm sure they have been tested well enough. > > Wolfgang > > Am 11.02.2011 00:42 schrieb "Wouter Hager" <ws...@gm...>: > > > Hi All, > > > > For some 4000 files, roughly 200MB, the xdb:store ant task I'm using has > > been running over 24 hours now. I am using eXist 1.4.1dev-rev13702 out of > > the box with 2GB of memory. and Lucene configured to use standard > analyzer > > only, with 4 qnames out of a dozen specified. I have set nodes-buffer to > > 1000. The files in the collection need to be overwritten. > > > > You probably heard it all before, but is there a way to speed things up a > > bit? The problem is that read access is slow while the store process is > > running. Am I doing something wrong perhaps? > > > > One final note: there are repeating elements in the docs... > > > > I noticed the Lucene buffer is set to 32MB. I read somewhere in the list > > that increasing it may speed up indexing a bit. This was in 1.4. Any idea > if > > this setting has effect in 1.4.1 and what is a good value? > > > > Thanks for any input! > > > > > > -- > > > > W.S. Hager > > Lagua Web Solutions > > http://lagua.nl > -- W.S. Hager Lagua Web Solutions http://lagua.nl |