I'm looking for a database that handles (fast, subsecond) queries on set-valued attributes. For instance, if I have a database of documents with tags, I'd like to be able to do a superset search. For input "tea,oolong,hot" I'd like to see:
- Hot Tea - tea,hot
- Oolong Tea - tea,oolong,hot
But not:
- Green Tea - tea,hot,green
For some realistic numbers, the number of documents would be approx 250k, the average number of tags would be 10-20, and the query might contain 100-1000 tags.
I've seen a lot of research papers in this area, but I haven't found anything concrete yet.
Though maybe not strictly an DBMS, Redis is very suited for this. See for instance:
http://rediscookbook.org/implement_tags_and_search_them.html
You can of course still store the rest of your data in any traditionaly DBMS