Question on MMap'd indices

[Sriram Sankar <sa...@gmail.com>]

In Unicorn (Facebook's search backend), we used mmap'd indices.  We could
load them on a separate process - which meant that we could make scoring
changes and test rapidly since we did not have to reload the index for
every run.  Is this true for Lucene also?  I'm assuming it would be if the
entire index content is within the mmap file and that there isn't and
additional bits of metadata sitting elsewhere.

Thanks,

Sriram.

RE: Question on MMap'd indices

[Uwe Schindler <uw...@thetaphi.de>]

Lucene is a library, it depends on your implementation how to keep index files open. Yes, Lucene supports mmap.

If you want to run Lucene as a separate server, use Apache Solr or ElasticSeach, which runs the indexing software as a separate process, accessible via REST API.

Uwe

-----
Uwe Schindler
H.-H.-Meier-Allee 63, D-28213 Bremen
http://www.thetaphi.de
eMail: uwe@thetaphi.de


> -----Original Message-----
> From: Sriram Sankar [mailto:sankar@gmail.com]
> Sent: Friday, June 21, 2013 6:19 PM
> To: java-user@lucene.apache.org
> Subject: Question on MMap'd indices
> 
> In Unicorn (Facebook's search backend), we used mmap'd indices.  We could
> load them on a separate process - which meant that we could make scoring
> changes and test rapidly since we did not have to reload the index for every
> run.  Is this true for Lucene also?  I'm assuming it would be if the entire index
> content is within the mmap file and that there isn't and additional bits of
> metadata sitting elsewhere.
> 
> Thanks,
> 
> Sriram.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org