-1

At the moment we have ASP.Net application with search based on Lucene.Net 3.0.3. And we are going to implement search service to work with > 2.5 mln items and have the similar questions, what search engine will be the quickest in this situation.

As we know Lucene.Net is based on classical Java Lucene principles and ideally it should have almost the same speed. But we found that Lucene.Net 3.0.3 engine have issues with speed of fuzzy search. We found some explanation why the performance is bad with fuzzy search in Lucene (on our data every request takes 6-8 seconds): Solr/Lucene fuzzy search too slow

Also our speed issues using Lucene.Net 3.0.3 are described here

So we have a list of questions for Lucene community and all experienced IT pro:

  • Does it makes sense to move from .Net to Java?
  • Do you see any other alternatives to work with such big amount of data?
  • Do you have such experience and can you share some numbers according to Lucene fuzzy search? (we had 4-8 sec per search request to 2.5 mln index, see link above to more details)
  • Do you have experience with FlexLucene? Is it better than Lucene.Net?

Thank you.

Community
  • 1
  • 1
SSoro
  • 1
  • 4

1 Answers1

0

I would suggest you to upgrade to a newer version of Lucene, as performance of Fuzzy Search has been improved significantly (by leveraging Finite State Machines).

Daniel Schneiter
  • 1,843
  • 1
  • 13
  • 19
  • Your advice makes a lot of sense , by where to move? Lucene.Net 4.8 isn`t released yet, FlexLucene is new and weakly documented( – SSoro Jan 14 '16 at 08:16
  • If Lucene (Java) is an option it's probably best to migrate to this one as you don't want to create a new migration dependency. – Daniel Schneiter Jan 14 '16 at 10:06
  • Thank you, I`ve made test with 1mln items and 30 fields: Custom .Net tool based on Lucene 5.4 is faster than Lucene 3.0.1 from 2-5 times for full text and wild-card search to 10-100 times for fuzzy search depending on complexity of requests. – SSoro Jan 19 '16 at 07:36