International Business Machines Corporation
Multiple parallel reducer types in a single map-reduce job

Abstract:

Different types of event data generated by a computing platform are transformed in parallel by a group of reducers during a reduce phase of a single map-reduce job into a set of different computed outputs that impose different types of storage formats based upon differing data characteristics of the respective different computed outputs. The set of different computed outputs are written to storage in parallel by the group of reducers during the reduce phase of the single map-reduce job according to the imposed different types of storage formats.