In this chapter we have designed a journey from a single core definition to multicore and then to distributed search. We discussed the analysis of multiple entities and how to choose between indexing them using different cores or denormalizing them into one single index.

A single, logical index could give us insights on how to best use our data (we left some suggestions to Carrot2 and similar tools, which we gave a fast introduction), permitting us to "see the full picture" of our domain.

Lastly, a logical index can be managed using SolrCloud; this way we will have it robust and scalable, as it relies on a collection of shards, and it is physically subdivided into many distributed cores.

Introducing these ideas is a further step toward a real system in production, but we need to take care of some fine tuning and customization, as we will see later.