S4 processes streams of data, one element at a time, as they arrive; outputs are produced incrementally. MapReduce and its derivatives (Cascading, Pig, etc) are

S4 processes streams of data, one element at a time, as they arrive; outputs are produced incrementally. MapReduce and its derivatives (Cascading, Pig, etc) are batch-oriented. Stream processing jobs can be massaged to fit into the MapReduce paradigm, but S4 provides a more natural solution.