elasticsearch vs solr - for custom full-text search systems

Question

I have a web application with a search engine created internally, based on Lucene indexes.

The application allows full text search.

The application is built in jsp and java.

To index user the Hadoop distributed processing software.

The Apache lucene project was customized to our specific case. Mainly the part of full-text search, to add new features of ordering results.

The front end uses hadoop-rpc to communicate with the backend, which in turn has created an internal solution (10 years ago) that researches in Lucene's indecisions.

The code is legacy, it needs a migration to another type of software. Here's the question: ElasticSearch vs. Solr ?

The requirements are:

• open-source, the purpose is to customize to accept specific features of the application's ranking.
• Working with Lucene indexes (since you do not want to reindex all over again).
• Speed are very important factors.

I have read some articles on the subject, but I wanted a personalized opinion for my case, any idea of which one will be the best option?

Thanks

Answer 1

SolR and Elasticsearch are quite similar. They both have the features you need.

It's a very open question and my opinion will be that Elasticsearch will be a better pick, it has a bigger community and tend to grow faster.

Answer 2

The more custom you need your solution to be, the more likely you will benefit from Solr.

While both of them use Lucene under the covers, Solr exposes more of it directly. Elasticsearch is tuned more for logs and things and some text search configuration is harder to access.