Our company produces large amounts of data - think financial data. We are building a system that requires us to read ~2T of data, perform complex floating-point calculations and write out about 100GB of transformed data.

There are 2 camps... C++ and Java at the company. Does anyone out there have papers (I've Googled already) that demonstrate Java's efficacy in data intensive, floating-point scenarios?

I should mention that this is a server-side application only - middleware - no GUI or client-side interaction required.

Thoughts?

Many thanks,
John