Python comes with built-in support for parallelising scripts locally over the compute cores of a single computer. However, as yet, there is no in-built support for parallelising scripts over the nodes of a cluster, e.g. distributed parallel programming.
Fortunately, there are several third party libraries that are beginning to appear that support distributed clusters, e.g. MPI4Py, IPython Parallel, Dask and Parallel Python. Most of these are based on the concepts of mapping, asynchronous functions, futures, and functional programming, so you should find that the concepts you have learned in parts 1 and 2 will be useful as you explore the developing ecosystem of distributed parallel Python libraries.
In this part, we will take a quick look at one such library, called MPI4Py.
MPI4Py is an actively developed third-party Python module that supports running parallel Python scripts across clouds, distributed compute clusters, HPC machines etc.
If you are running Python on your own computer, then you can install MPI4Py using either:
pip install mpi4pyif you have installed
conda install -c anaconda mpi4pyif you are using Anaconda Python or use the Environments tab in Anaconda Navigator.
Note that in order to install this library though
pip, you will need to install the MPI libraries as well. Anaconda can install them for you automatically.
For more information regarding installing the library on your own computer, see the official documentation.
If you are using Python installed on a cluster, then MPI4Py should already have been installed for you (if not, email your cluster systems administrator).