Re: Interesting, but
I second your article suggestion, this would be really interesting. In my previous job at a consulting firm I went on a two day MongoDB course as they were trying to build a "big data" capability without really knowing why. I found the course interesting but at the end I came away from it not that sure why I would want to use it.
The main advantage I could see was the ease of scaling horizontally with sharding as doing this type of thing is very, very difficult with a conventional RDBMS. However, I've also never encountered a situation where "eventual consistency" was deemed acceptable.
Also, with a well designed and well indexed SQL database you can handle "moderately big" data fairly well. In my current job I managed a data warehouse built using £15k of commodity wintel hardware, SQL Server 2008R2 and fast SAN storage, and we can run queries across hundreds of millions of rows of data and joining numerous tables in seconds.
What would be fascinating is to see how NoSQL could handle "properly big" data and how people get around the consistency issues. For example, I would love to import the full event stream from our Adobe SiteCatalyst instance and query it - is this the kind of thing I can turn to NoSQL for?