NoSQL and Traditional DBs
The term NoSQL is not a very well-defined term.
There has been a rebirth of interest in database technology in the past 10 years. This rebirth has been sparked from an unexpected source — not from the core Database community, but from the Distributed Systems community.
People working in large-scale, high traffic, distributed, internet businesses such as Akamai, Google, Yahoo, and Amazon have made major strides in distributed data stores. Out of a necessity to handle high-volume traffic and to always be up, these business hired distributed systems Ph.Ds. and invested in non-traditional database technology.
The data stores that came from these roots are what we tend to refer to as NoSQL. They are no relational or OLAP.
Since they do not hail from transitional-DB roots, they do not need to build on a base established in the 70s. They are free of the baggage (or so we try) of relational and OLAP DB vendor comparisons.
They can be evaluated on their own merits and features. In fact, most of them are as different from one another as they are from relational databases.
Hence, when selecting a NoSQL DB, keep an open mind, forget what you learned about relational and OLAP DBs. Know what features you need and look for the engine the provides it.
P.S. : This comment came from my response to the following question on Quora