This workflow diagram from a Neuroinformatics paper describing the Extensible Neuroimaging Archive Toolkit (XNAT) is a perfect illustration of how I think storage, collaboration and public release of scientific data should work. Ideally funding bodies (or some other overarching organisation) would supply this kind of infrastructure for all their projects, based on S3 and EC2 or similar.
Brainscape, built on XNAT, is a finalist in the Amazon Web Services Startup Challenge: