Design, Implementation & Performance Analysis of Low Cost High Performance Computing (HPC) Clusters

Abstract

More accuracy and minimizing time needed for the scientific calculations in the field of computing has become an important task currently. Scientific problems are computationally intensive and a lot of time is acquired to get the outcomes by utilizing computer with one processing element. High Performance Computing (HPC) describes utilization of many processors executes a computationally intensive task in parallel, which reduces the processing time. At present, one of the leading and fastest parallel computers is supercomputer contains the processing elements hundreds to thousands. A lot of effort and cost is needed to build very fast and dedicated supercomputers. Alternative is to build low cost HPC clusters using inexpensive hardware. In this paper, we implemented tested, compared and analyzed the processing ability of 3 different x86-based and ARM-based low cost HPC clusters in terms of execution time and speedups. Finally concluded that in spite of the performance, clusters build with ARM-based single board computers are functional and handy for academic point of view due to its low-price and value to the restrictions of electricity.