Really this is based on my readings and few experiments, I'm not 100% sure that datalog can not be made faster and socialite (see below) seem to claim I'm wrong.

Datalog is the native query language of datomic. I can put the finger on it, but I think there is a project for querying cassandra with datalog+clojure.

This is the most recent work I'm aware of:

- python querying of hadoop with a datalog-like language http://socialite-lang.github.io/

- overview: http://fr.slideshare.net/Hadoop_Summit/t-325p211seo

Also best way to learn datalog: http://www.learndatalogtoday.org/