I've been fascinated with Key-value databases for almost as long as my career. Interestingly enough, I first came across the concept when I did a little work with Lotus Notes, way back in 2001. My memory is a little rusty on the full capabilities of Lotus databases, so I don't know if it applies to their format, but what I miss most from other HashMaps in the cloud are aggregate functions.

As a developer, I'd prefer one of two innovations: relational databases that scale better across machines or k/v databases that provide a familiar mechanism for ad-hoc querying and aggregation across keys.