I built an in-memory NoSQL index (not a database) for answering subset inclusion queries, i.e. "find me all the Items with {Feature1, Feature2, Feature3}". It also supports queries by category, and by color.
I say it's not a database because it isn't even a Key-Value store. It's just a key-store - you query it for the keys, and retrieve the values from somewhere else (e.g. postgres).
It's vastly faster than SQL for this purpose.
http://www.chrisstucchio.com/blog/2012/introducing_hobo.html
https://github.com/stucchio/Hobo
I say it's not a database because it isn't even a Key-Value store. It's just a key-store - you query it for the keys, and retrieve the values from somewhere else (e.g. postgres).