Interesting paper (and makes me feel more justified in rejecting Hadoop). Do you have any blog posts / other material about the techniques you're using at Sense?