Keen just stores blobs in Cassandra and does aggregation at query time. I can't imagine it would scale far enough to be useful in this situation.