Hadoop isn't just extended by additional functionality, there are tools to provide entirely different paradigms for writing the code used to process your data within Hadoop.
Developed by Concurrent, and open sourced under an Apache license, Cascading (http://www.cascading.org/) is a popular framework that abstracts the complexity of MapReduce away and allows us to create complex workflows on top of Hadoop. Cascading jobs can compile to, and be executed on, MapReduce, Tez, and Spark. Conceptually, the framework is similar to Apache Crunch, covered in Chapter 9, Making Development Easier, though practically there are differences in terms of data abstractions and end goals. Cascading adopts a tuple data model (similar to Pig) rather than arbitrary objects, and encourages the user to rely on a higher level DSL, powerful built-in types, and tools to manipulate data.
Put in simple terms, Cascading is to PigLatin and HiveQL what Crunch is to a user-defined...