Many systems have provided restricted programmingHas anyone implemented parallel prefix scan as an extension to MapReduce or a similar framework such as Hadoop, and did it prove useful?
models and used the restrictions to parallelize the computation
automatically. For example, an associative function
can be computed over all prefixes of an N element
array in log N time on N processors using parallel prefix
computations.
Did anybody do parallel prefix using map reduce yet?
ReplyDelete