Parsl is a parallel scripting library for Python that allows users to develop and execute dataflow scripts on parallel and distributed resources.
Parsl runs programs concurrently as soon as their inputs are available, reducing the need for complex parallel programming. Parsl expresses dataflow in a portable fashion: the same script can run on multicore computers, clusters, clouds, grids, and supercomputers.
- This repository includes three tutorials:
- parsl-introduction.ipynb: a quickstart guide to using Parsl
- parsl-workflows.ipynb: example workflow patterns expressed in Parsl
- parsl-advanced-features.ipynb: examples of various Parsl features, such as multi-site, elasticity, and fault tolerance
The tutorial repository includes everything needed to run these notebooks. It includes sample applications simulate and stats (mock "science" applications) as well as a basic MPI program. It also includes configuration files for several compute resources.
The easiest way to run these tutorials is via Binder. Using Binder you can run the tutorial notebooks in your browser without installing any code locally.
To run the tutorial locally, ensure that Python (3.5+) and Parsl are installed.
To install Parsl:
$ pip3 install parsl
To install Parsl from source, see the documentation.
Finally, set up the Parsl tutorial:
$ git clone https://github.com/parsl/parsl-tutorial.git