Raspberry Datacube Cluster

A Raspberry Pi is a complete single-board computer, lightweight yet versatile. The SkyFed project proves that devices of such size and make are suitable for on-board datacube processing. To this end, the rasdaman datacube engine got ported to the Raspberry hardware.

A Raspberry Pi is a complete single-board computer, lightweight yet versatile. We prove that devices of such size and make are suitable for on-board datacube processing. To this end, early in the project the rasdaman datacube engine got ported to the Raspberry hardware.

A custom cluster of 64 Raspberry Pi 5 was built (largest ever so far), connected through Wifi (cables seen in the picture are power supplies). An ERA5 climate datacube was distributed manually over the nodes, grouped into a single virtual datacube. Without any coding (such as Python), WCPS geo datacube queries now allow distributed processing with automatically balanced workload. In a next step, existing large-scale archives were federated: the CoperniCUBE datacube service and the aviation weather service established in NATO SPS Cube4EnvSec.

In the first public presentation at Space Tech Expo in November 2025, the expert audience were deeply impressed by the new opportunities unfolding.

The videos below describe how the cluster was built (left) and explain the strategic impact of this innovation (right). This is also summarized in a GEObusiness article (German version).