I've found that the techy types at the NOSQL vendors are usually pretty honest about what their use case and what doesn't fit their use case.

The marketing types are another matter entirely. One marketing type extolled the virtues of their distro of Hadoop stating that they could ingest 500GB in half an hour and that was way beyond the capabilities of traditional RDBMS. Clearly they hadn't seen the Microsoft article on loading 1TB in 30 minutes.

One of the SQLCat labs at SQL Bits (Liverpool) actually went into the details of how this was done and it was remarkably straight forward.

Of course the Hadoop marketeer didn't give details of what that 500GB looked like. Was it a simple log file or a complex structured JSON document. we'll never know!

Then there is the irritating tendency to demonstrate the NOSQL products superiority by using an awful schema in an RDBMS to say that the RDBMS can't cope. That's like saying that if you make Usain Bolt wear diving boots it prooves he's a lousy sprinter.