[Xapian-discuss] Xapian and Solr comparison

[Olly Betts]

On Fri, Mar 13, 2009 at 05:11:46PM +0000, Charlie Hull wrote:
> Anurag Goel wrote:
> > I have been evaluating Xapian and Solr/Lucene for building a search
> > application, and ran some tests to compare indexing and query performance
> > between the two.
> > 
> > I think the results might be useful to others, and I have posted them here:
> > http://www.anur.ag/blog/2009/03/xapian-and-solr/
> > 
> > In a nutshell, Xapian's index size (using flint) was much larger than
> > Solr's, but the query performance was much better.
> 
> Just noticed another comparison, this time from 2006:
> http://grupoweb.upf.es/WRG/dctos/Middleton-Baeza.pdf

Three-year-old comparisons are close to worthless.  Xapian has changed
dramatically in that time, and I bet several of the other packages in
their comparison have too.

> Not sure if I'd seen this before. Unfortunately for some reason Omega 
> (or Omindex) failed to index the test collection in a reasonable time, 
> so was excluded from the final tests.

It says they were using Omega 0.9.5, and presumably the same version of
xapian-core.  There have been quite a few improvements since then...

It also sounds like they would have done better to set
XAPIAN_FLUSH_THRESHOLD higher (the machine had 2GB of RAM).  It would be
good if we could tune this automatically (or at least set the default
depending on the machine) as it does tend to work against us in this
sort of "black box" benchmarking.

Cheers,
    Olly