This is something that I struggled with when I ran my technical spike. It is A solution, not necessarily THE solution to a number of problems.

The case study that Thoughtworks quote for Autotrader is where millions of PDF documents had to be scanned and relevant facts extracted for the Autotrader web site.

For mining web log files then a dedicated solution such as Tibco Log Logic or Splunk are probably a more targetted solution.

Unless something changes dramatically in the Hadoop architecture I don't see it as being a serious datawarehouse alternative. It is at its heart a file scanning tool. It isn't designed for multi-concurrency, random access, security etc.

I think there is a danger that people think they have to jump onto the Big Data bandwaggon when the reality is that few people actually have genuine Big Data problems. They probably have people, process and politics problems causing technical problems