Dask is a project of Continuum Analytics (the same company that's responsible for Numba and the conda package manager) and a pure Python library for parallel and distributed computation. It excels at performing data analysis tasks and is very well integrated in the Python ecosystem.
Dask was initially conceived as a package for bigger-than-memory calculations on a single machine. Recently, with the Dask Distributed project, its code has been adapted to execute tasks on a cluster with excellent performance and fault-tolerance capabilities. It supports MapReduce-style tasks as well as complex numerical algorithms.
