I did not quite understand this part of the lecture,but although it was just a brief walk through, I still see the power of its algorithm.
I thought this was a good picture that taught me the basics of it, and I hope it would help others who also have doubts on it.
An example given was that the Google search engine uses a lot of MapReduce to lower their runtime costs. I can imagine the work of processing so much data. The basic steps taken are:
1. Read the raw data from the storage and pass it onto the next step.
2. Filter the data through a map function that returns a collection of pairs of names and values. If the data is to big, then it should be separated into sub-sets of data and be ran on parallel map functions onto the next step.
3. Shuffle these pairs of names and values together and group those data that have similar charisteristics.
4. For each group, run them through a reduction function that returns collections of values
5. store the final sets of values for use.