High-performance computing: A cost effective and energy efficient approach

Abstract

The world is witnessing unprecedented advancements in ICT (Information & Communication Technology) related fields. These advancements are further boosted with the emergence of big data. It goes without saying that big data requires two major operations: storage and processing. The latter is usually provided through High-Performance Computing (HPC) which is delivered through two main venues: supercomputers or clustering. The second venue has been widely opted for as a cost-effective alternative when compared to supercomputers. However, with the widespread increase in deploying ICT-based applications, the parallel increase in energy consumption has become a real issue. Thus, researchers have been exploring approaches to conceive big data analytics platforms that are both cost-effective and energy-efficient. In this paper, we present a cost-effective and energy-efficient HPC clustering that is based on Raspberry PIs. Our approach leverages the concept of Green Computing. We evaluated our cluster performance and its energy consumption and compared it to a commodity server. We leveraged on the Amdahl's law to set the maximum speedup of the proposed approach. Our approach can be easily deployed for usage in different ICT-based applications that consider energy efficiency as a priority.