“NoSQL” tools throw Data Consistency out completely and maximizes Performance by eliminating relational constructs such as schemas and normalization. Great for some applications, horrible for others.

For now, I stick to relational/SQL unless I’m absolutely convinced that there is no way to achieve my objectives with it. I am building a GIS iPhone app with an embedded SQLite back-end because it has built-in R-Tree indexing. A decade ago I wrote a text search engine with a BerkeleyDB back-end because all relational databases I tried sucked at maintaining an inverted index of millions of terms and running cosine co-efficient calculations of query/document vectors. If I had to re-write this today, I would probably take a good look at Mongo and Couch.