Very similar to style #30, but with an additional twist
Constraints:
-
Input data is divided in chunks, similar to what an inverse multiplexer does to input signals
-
A map function applies a given worker function to each chunk of data, potentially in parallel
-
The results of the many worker functions are reshuffled in a way that allows for the reduce step to be also parallelized
-
The reshuffled chunks of data are given as input to a second map function that takes a reducible function as input
Possible names:
- Map-reduce
- Hadoop style
- Double inverse multiplexer