I'm pretty sure that HADOOP
is exactly what you want, and that map / reduce
is the key to making sense of the data in its pure (non-semantically enhanced) form.
Big Data | Microsoft SQL Server
already delivers the ability to do what I think that you want... Allowing you to add semantic meaning via relational data merged with the HADOOP.
Is this what you're driving toward? If not, please try to explain what I'm missing.