Configuring Hadoop with Cassandra in itself is quite some work. Writing verbose and long Java code to do something as trivial as word count is a turn-off to a high-level user like a data analyst. Wouldn't it be nice if we had a SQL-like interpreter that converts commands to MapReduce programs for us? Pig is exactly that tool.
Note
Hadoop does not only support Java but also MapReduce programs can be written more concisely in multiple languages such as Scala, Python, C++ (Pipes), R, and many adapter languages.
Pig provides a SQL-like language called Pig Latin. One can write complex MapReduce programs using Pig Latin. You can create a set of intermediate variables that are a result of an operation and it can be used in subsequent operations, in the same way a stored procedure in RDBMS would. Finally, the output of an operation can be displayed on a screen or can be stored in a permanent storage such as HDFS or Cassandra.