There are two general viable use-cases for noSQL:
- Rapid application development
- Massively scalable systems
The fact that most noSQL solutions are effectively schema-less; require far less ceremony to operate; are light-weight (in terms of API); and provide significant performance gains in contrast to the more canonical relational persistence systems informs their suitability for the above 2 use-cases (in the general sense).
Being cynical -- or perhaps practical in the business sense -- one can propose a 3rd general use-case for noSQL systems (still informed by the above set of characteristics/features):
It is easier to grock and any inexperienced (but un-brain-dead) aspring geek can pick it up in a snap. That is a very powerful feature. (Try that with Oracle ..)
So, the use-cases of noSQL systems -- which in general can be characterized as relaxed persistent systems -- are all optimally informed by practical considerations.
There is absolutely no question -- outside of hugely massively scalable systems -- that RDBMS systems are formally perfect systems designed to insure data integrity.
