Supercomputing has become an integral part in many applications and fields of research. Nowadays, the most powerful machines are so called distributed memory systems. This means that there is a large number of compute nodes, each having its own memory, where the interconnect is realized with some kind of network. It is thus important to understand how to design algorithms in order to scale with the number of cores in such a machine. For this purpose, we built a cluster of Raspberry Pis reflecting the architecture of todays most powerful computers at an affordable price. In this talk we present the building instructions from both a hardware and software perspective. We also discuss some results that were achieved with this machine in a tutorial for students of mathematics at Trier University and why this understanding is important for the design of future algorithms.

Von Martin Siebenborn