00:01
The mapreduce paradigm is a programming model for distributed computing that's well suited to large -scale data processing tasks in a utility cloud.
00:11
In the mapreduce paradigm, a computation is divided into two stages, the map stage and reduce stage.
00:28
During the map stage, a set of input data is divided into smaller chunks, and a mapping functions apply to each chunk to reduce a set of intermediate results.
00:38
During the reduced stage, these intermediate results are combined using a reducing function to produce the final output of the computation.
00:48
The map reduced paradigm is well suited to distributed algorithms that involve the processing of large amounts of data, such as those in data mining, machine learning, and scientific computing.
01:00
It's particularly effective for tasks that can be parallelized, such as the computation of large matrix multiplications, or the analysis of large data sets.
01:11
But it's not well suited to algorithms that require fine -grained control over the execution of individual tasks, or that require frequent communication between tasks.
01:21
These types of algorithms may be better suited to other programming models, such as mpi, which is message passing interface, or shared memory programming.
01:31
Here are some examples.
01:34
First examples that are well suited to the map, reduce paradigm.
01:38
This would be, i'll note that these are well suited.
01:49
Word count.
01:53
This is a simple example of a map produced computation in which the input data is a large collection of documents and the output is a count of the frequency of each word in the documents...