Performance Studies of 30-Node (120 Core) Raspberry Pi4 Beowulf Cluster

Raspberry Pi is an emerging technology that can be used to build cluster computers for conducting computational tasks that can be built at a fraction of the cost compared to the conventional rack-mounted cluster computers. We have built a portable table-top 30-node Raspberry Pi4 Beowulf cluster to explore the capabilities of Raspberry Pis to determine how well the Raspberry Pis perform in a cluster computing environment for parallel computational tasks. The Raspberry Pi4 cluster uses PoE to deliver power to each node. The Raspberry Pi4 cluster consists of 120 cores; each node has a quad-core 64-bit Cortex-A72 (ARM v8) CPU connected to a 48-port gigabit ethernet switch and is running the 32bit Raspbian-11 operating system. We have tested the performance of ARM (Advanced RISC Machine) CPUs. We have compared the performance of the 30-node Orchard Raspberry Pi4 cluster with a previously built 33-node Raspberry-Pi (RPi) cluster with 33 cores, which has Raspberry Pi1 Model B+ CPUs running the Raspbian-7 operating system. Both Raspberry Pi clusters use OpenMPI for parallelization across the nodes. We ran three MPI codes using the OpenMPI library for the computational tasks and to conduct benchmark tests. We will present the results that show the performance across the cluster nodes.