Cubed: Scalable array processing with bounded memory

Cubed is a Python library for scalable out-of-core multi-dimensional array processing with bounded memory.

Familiar API

Cubed provides NumPy and Xarray APIs for processing your multi-dimensional array data

Dask replacement

Cubed is a drop-in replacement for Dask’s Array API

Predictable memory usage

Cubed will tell you if your computation would run out of memory before running it

Reliable

Cubed is designed to be robust to failures and will reliably complete a computation

Run locally

Cubed can process hundreds of GB of array data on your laptop using all available cores

Scale in the cloud

Cubed is horizontally scalable and stateless, and can scale to multi-TB datasets in the cloud