From the webpage:
Peregrine is a map reduce framework designed for running iterative jobs across partitions of data. Peregrine is designed to be FAST for executing map reduce jobs by supporting a number of optimizations and features not present in other map reduce frameworks.
Among its many “modern” features, Peregrine includes: “MapReduceMerge style computations including a new merge() operation.”
I will have a separate blog entry on a paper describing MapReduceMerge computations for heterogeneous data sets.
This looks very important for the future of topic maps in a big data (heterogeneous) universe.