RDBMS is not king
The NoSQL 'movement' is a reaction against the mantra that relational databases are the best tool for the job of storing data. This is a very subjective statement, and one open to a lot of valid criticism, as the you cannot be so certain that they are suitable for _every_ job.
I think the reaction is a good one, there are horses for courses. RDBMS' of various flavours are extremely powerful tools, and are stuitable for many, or maybe most data storage jobs. They are not suitable for everything, however.
I'm bored of implemeting versioned data, for example. It can be done in an rdbms, of course it can, but the answer is, should it? Is there not a better model? The joins with multiple versioned data sets can be horrendous, and slow!
On the comment above on data consistency, this is a very valid concern. the no sql databases are still quite young, and there are implications of the models they promote are not yet fully understood.
An example from my work where an rdbms was the wrong answer.
I was implementing a message matching system (SMS). Message goes out, and delivery receipt comes back from the network. This needed to be matched against the original message id and the message marked as delivered.
Now, this was implemented in an oracle db. It was OK, we were getting up to 30 matches a second during peak load. This was the major limit in the scalability of the system.
Instead of this, we switched to a model where the outgoing message id was written a file, sorted using GNU sort for a simple sequential match. Every 5 seconds or so the incoming receipts were batch matched against this file. Potential throughput went to multiple thousands per second.
The results were then batch updated every minute or so into the DB for reporting purposes.
So, in summary, the DB was great for reporting and maintaining data. For the match use case, it was a drag, the wrong tool for the job.
This is a_very_ simple example; but then, I'm trying to make the case that you should use the right tool for the job, and a relational database is not always that tool.
I'm not against them, by any means, they're very good at what they do, but so is gnu sort, a file system and a bit of groovy scripting.