This is good example of how to answer 'big data' questions without big expensive distributed systems via the magic of lots of RAM and probabilistic data structures.