Designing a High Performance Parallel Personal Cluster

TL;DR

This paper presents a low-energy, high-performance personal computing cluster using single board computers based on the Beowulf concept, aimed at scientific and engineering applications with cost and energy efficiency.

Contribution

It introduces a novel low-energy architecture with SBCs for scientific computing, demonstrating performance and reliability through benchmarking and real-world tests.

Findings

Achieved competitive performance with low power consumption.

Validated reliability with practical industry benchmarks.

Provided a cost-effective alternative for scientific computing.

Abstract

Today, many scientific and engineering areas require high performance computing to perform computationally intensive experiments. For example, many advances in transport phenomena, thermodynamics, material properties, computational chemistry and physics are possible only because of the availability of such large scale computing infrastructures. Yet many challenges are still open. The cost of energy consumption, cooling, competition for resources have been some of the reasons why the scientific and engineering communities are turning their interests to the possibility of implementing energy-efficient servers utilizing low-power CPUs for computing-intensive tasks. In this paper we introduce a novel approach, which was recently presented at Linux Conference Europe 2015, based on the Beowulf concept and utilizing single board computers (SBC). We present a low-energy consumption architecture capable to tackle heavily demanding scientific computational problems. Additionally, our goal is to provide a low cost personal solution for scientists and engineers. In order to evaluate the performance of the proposed architecture we ran several standard benchmarking tests. Furthermore, we assess the reliability of the machine in real life situations by performing two benchmark tools involving practical TCAD for physicist and engineers in the semiconductor industry.