Cost-Effective Clusters in Education: A Model for Practical HPC Learning

This innovative practice paper details a teaching experiment utilizing a cluster that simulates a real HPC en-vironment. Insper, a new higher education institution in the engineering field, aims to provide practical and interactive education. To this end, a course was implemented for 40 seventh-semester Computer Engineering students. The course explored the use of a low-cost cluster inspired by supercomputers from the TOP500 list, including Frontier in the United States, Fugaku in Japan, and Santos Dumont in Brazil. This research focuses on enhancing the HPC module within the Computer Engineering program by integrating practical HPC experiences using a cluster and assessing its educational impact through both quantitative and qualitative measures. Data were collected from system logs and a structured questionnaire admin-istered to 40 participating students. Analysis of the logs provided insights into job submissions, resource utilization, and common errors encountered during cluster operations. The questionnaire assessed students' understanding of HPC concepts, their practical skills, and overall satisfaction with the course enhancements. Results indicated a significant improvement in the students' practical understanding of HPC concepts, as evidenced by the successful completion of 177 jobs and enhanced student engage-ment with the practical components of the course. However, chal-lenges such as resource allocation errors and variations in system performance highlighted the need for ongoing improvements in both educational strategies and cluster infrastructure. The study concludes that integrating a hands-on HPC clus-ter within the curriculum significantly enhances the learning experience by making complex concepts such as parallelism and resource distribution more tangible. Future recommendations include focusing on reducing system cluster errors, improving resource estimation education, and standardizing the system to ensure consistent student experiences. This approach not only facilitates the application of theoretical knowledge but also prepares students for real-world computational challenges.