lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "SK R" <rsk....@gmail.com>
Subject Re: speedup indexing
Date Tue, 07 Aug 2007 10:23:36 GMT
Hi,
    Thanks for this valuable informations.
    I'm using Lucene2.1 now. Do I need to apply the patch  "LUCENE-843" with
existing one or i have to move the latest? Do i need to use flushByRam
instead of flushbydoc to work with this patch?


Regards
RSK

On 8/7/07, Michael McCandless <lucene@mikemccandless.com> wrote:
>
>
> "Mike Klaas" <mike.klaas@gmail.com> wrote:
>
> > > On 8/6/07, testn <test1@doramail.com> wrote:
> > >>
> > >> 2. To improve indexing speed, you can consider using the trunk
> > >> code which
> > >> includes LUCENE-843. The indexing speed will be faster by almost
> > >> an order of
> > >> magnitude.
> >
> > While a speedup should be expected, I don't know that an order of
> > magnitude is a realistic expectation to convey.  Unless, of course,
> > you're speaking in base two ;)
>
> Right, it's important to not overstate things here...
>
> First off, the speedups in LUCENE-843 *only* apply to the actual time
> spent in Lucene's indexing code.  Ie, time spent retrieving the doc,
> running the analyzer, etc., will not get any faster (though there is
> ongoing work to speed up the core analyzers...) so if the bulk of the
> time in an application is not in Lucene's indexing then the speedups
> will be minor.
>
> Second off, the speedups are best for smaller docs, and, at present
> you still need to either change your writer to flush by RAM or set a
> large "maxBufferedDocs" in order to see the best gains.
>
> It's also important to try these suggestions too -- they can
> potentially make even more difference than LUCENE-843:
>
>   http://wiki.apache.org/lucene-java/ImproveIndexingSpeed
>
> Mike
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message