You should look into Spark: http://spark-project.org/ and Shark (Spark + Hive).

I personally find that using the Spark repl is both efficient and flexible; and being able to cache data in memory makes for very fast iterations, especially when doing something more sophisticated.

There's also Cloudera Impala, but I've never used it, nor do I think I'll ever have to given what the AMPLab folks are doing.