Dask

Dask is an open-source Python library designed to natively scale the Python scientific stack (like NumPy, Pandas, and Scikit-learn) to handle datasets larger than memory or to execute computations across distributed clusters. It achieves parallel computing through dynamic task scheduling,…

Key facts

First appeared
2014
Category
technology
Problem solved
Dask was created to address the 'memory wall' and 'single-core wall' faced by Python data scientists. While libraries like NumPy and Pandas excelled for in-memory, single-machine computations, they struggled when datasets grew beyond RAM or when extensive parallel processing was required, forcing users to either downsample data, rewrite code in JVM-based languages (like Scala for Spark), or manage complex custom parallel scripts.
Platforms
macOS, Linux, Cloud computing environments (AWS, Azure, GCP), Kubernetes, HPC clusters (via schedulers like SLURM, PBS), Windows

Related technologies

Notable users

  • Capital One
  • Pangeo Project
  • Saturn Cloud
  • Los Alamos National Laboratory
  • Google
  • Coiled
  • NVIDIA