Calculating matrixes in parallel code

• Main Author: Κλαδής, Διονύσης
• Other Authors: Kladis, Dionysis
• Language: English
• Published: 2022
• Subjects: High performance computing (HPC); Hypervisors; Υπολογιστές υψηλής απόδοσης; Εικονικές μηχανές

This material focuses on the question of whether there is currently the necessary architectural support, provided by hardware for the efficient execution of virtual machines used in High Performance Computing (HPC). The High Performance Computing (HPC) technique has evolved as there has been an increase in computing power demand. This is related to the development of technology and complex research work that leads to an increase in the need for rapid calculations. The technique of High Performance Computing (HPC) is increasingly used in all kinds of fields of science and beyond. Today, in all aspects of science, it takes a huge amount of time to solve scientific problems. So sustainable computer supercomputers, on the other hand, are very expensive to build and monitor. Virtual machine technology is a field with ongoing research even today. Despite the industry’s current focus on architectural support on current software and hardware platforms, we believe that some historical perspective is needed to address the problem properly. The first half of this paper provides a historical perspective and theoretical framework developed five decades ago by major leading companies in the field. It also describes older systems and applications that defined modern virtual machine development and High Performance Computing (HPC) techniques. As is often the case in active and rapidly evolving fields such as information technology, theory defines some necessary but not sufficient features. Therefore current architecture is the result of combining the theoretical framework with the ideas derived from real applications in these systems. The second half of the paper describes the steps we have taken to utilize the technology of virtual machines relying on the technique of High Performance Computing (HPC), with the specific way of implementing, where the Hypervisors act as graphical terminals for one of the virtual operating systems for possible use in common areas and production scenarios. This material presents a such use case on our implemented infrastructure, we have created in python a matrix calculation code in parrallel and in sequential modes and executing them through in our local computer and at the cluster. It compares performance times of execution times through a test code for the Cluster test of High Performance Computing (HPC) virtual machines. In conclusion, the performance of the computer farm platform is checked by comparing the execution times and data processing speeds in High Performance Computing (HPC) Cluster when utilizing the full capabilities of processors in Hypervisors. Finally, the process and the data collected are reviewed and concluded by reviewing the results and proposing possible improvements.