µ-Cluster: Everything Parallel

Funding: University of Basel

Duration: 01.03.2017-Present

Project Summary
The μ-Cluster is a micro-parallel computing cluster consisting of 66 single-board computers (SBCs). Originally developed using Odroid-C2 devices, the system underwent a major hardware upgrade in Fall 2025 and is now based on Raspberry Pi 5 SBCs. Each node features a quad-core ARM Cortex-A76 processor running at 2.4 GHz. One SBC serves as a login node, while another acts as a central storage node for the remaining 64 compute nodes.

The 64 compute nodes are arranged across four separate levels, with 16 SBCs per level, inside a cubic-shaped transparent acrylic enclosure. Each of the four vertical facets of the case contains 16 LCD touch screens, one display per compute node, connected to the SBCs located on the corresponding level. The base of the enclosure houses the power supplies. The SBCs are interconnected in a star topology using 22 Netgear 1 Gbit/s Ethernet switches. In addition, each compute node is coupled with an Intel Neural Compute Stick (NCS), enabling efficient neural network inference workloads.

The primary goal of the μ-Cluster project is to “Visualize occurring parallelism in real time!”. The system is designed as a portable and visually appealing platform for audiences unfamiliar with parallel and distributed computing. At the same time, it serves research and teaching purposes in areas including high-performance computing, operating systems, databases, networking, information retrieval, machine learning, and systems engineering. The μ-Cluster is used for performance analysis and benchmarking of CPU-intensive, memory-intensive, and network-intensive workloads, as well as experimentation with distributed and heterogeneous computing systems. In addition, the system is regularly showcased during public outreach events, including Fantasy Basel and the Bachelor Infotag at the University of Basel, where it is used to demonstrate concepts of parallel and distributed computing to a broader audience.