A Simple Performance Comparison

I've compared the indexing performance of Montezuma against that of Lucene and Ferret. This is an extension of the work Dave Balmain did comparing Ferret and Lucene, and I used the same corpus (Reuters 21578) and his code for Lucene and Ferret.

The test is to index 19,043 Reuters articles using the simple whitespace-analyzer. Term vectors and the documents themselves are not stored.

	Montezuma 0.1.1	Montezuma 0.1.2	Lucene 2.0	Ferret 0.10.13
documents/s	31.62	41.74	56.01	521.6

I used Red Hat Fedora Core 5 (FC5) running under VMWare in Windows XP to do the comparisons. SBCL 0.9.18 was used for Montezuma.