An\'alisis de rendimiento y eficiencia energ\'etica en el cluster Raspberry Pi Cronos

TL;DR

This study evaluates the performance and energy efficiency of a Raspberry Pi-based cluster for educational use, analyzing scalability, stability, and power consumption with benchmark tests.

Contribution

It provides a detailed assessment of a low-cost ARM cluster's performance and energy metrics, highlighting the impact of node heterogeneity.

Findings

Cluster achieves up to 6.91 GFLOPS with homogeneous nodes

Heterogeneous nodes reduce stability and efficiency

Electrical consumption measured for performance-to-power ratio

Abstract

This article presents an evaluation of the computational performance and energy efficiency of the Cronos cluster, composed of Raspberry Pi4 and 3b microcomputers designed for educational purposes. Experimental tests were performed using the High Performance Linpack (HPL) benchmark, under a resource management environment configured with Slurm and parallel communication via Open MPI. The study focuses on analyzing scalability, stability, and power consumption during the execution of computationally intensive workloads, considering different node configurations. The results show that the cluster achieves a performance of up to 6.91 GFLOPS in homogeneous configurations of 6 Raspberry Pi 4 nodes, and that the use of heterogeneous nodes (including Raspberry Pi 3b) can negatively impact stability and efficiency. Additionally, the total electrical consumption of the system was measured during the runs, allowing for the estimation of the performance-to-consumption ratio (GFLOPS/W) as a comparative metric. This study constitutes a concrete contribution to the design, evaluation, and utilization of low-cost ARM clusters in educational and research contexts.