It's not that relational databases aren't good at what they do nor is it that developers don't understand them. They're fantastic at it; it's just what they do

It's not that relational databases aren't good at what they do nor is it that developers don't understand them. They're fantastic at it; it's just what they do isn't what app developers really need. For the most part, app developers aren't looking for a store that allows them to query their data 6 ways from Sunday if they write 50%+ more code to put their data in there. What they're looking for is rock solid transactional persistence for the data structures their applications actually use. Relational databases primary use in the enterprise market is really as an integration point, a central location for all business data that all applications can access. It does this very well, but the times are changing and more and more apps are being written Internet style where integration is done via web services at the application level. Relational databases aren't going anywhere any time soon, but their dominance as the default solution of choice is coming to and end. Object/Graph/Key-Value database are being used more and more because they actually solve the problem app developers want solved; they reduce the code that needs to be written in the application to deal with persistence to virtually nothing. Developers are perfectly willing to trade queryability and and data consistency for simplicity, availability, and easy scalability across farms of servers. Dealing with the consequences of pointer chasing is less work that writing the code necessary to fit a graph of objects into a relational database.