Communication Patterns in Cloud Haskell

I'm really excited to try Cloud Haskell out. I've been working with Spark for a while and, while it's powerful, the Hadoop/Java legacy is a pain point, and keeping Java-culture out of Scala seems impossible (if leading open-source libraries can't do it, then what hope is there?)

I'd love to see a point where people don't assume that distributed programming automatically means Hadoop, because I feel like (a) those solutions are heavyweight and user-hostile due to Java inheritance and (b) Hadoop and Spark are the only things propping the JVM up, these days. (It helps that Scala and Clojure are half-decent languages, although I've grown to prefer Haskell.)