An Efficient Framework for Enabling Virtual High Performance Computing Clusters on Demand in Cloud

Abstract

To handle difficult computational issues, scientists and researchers have extensively utilized high-performance computing (HPC) equipment in both commercial and academic institutions. Most intricate computer issues typically involve either high volumes of data or intense computational processes. Addressing these issues may necessitate extended periods, ranging from hours to days or even weeks, for their successful execution. For instance, certain calculations in traditional HPC systems take weeks to complete and need 1,00,000 processors. Therefore, traditional HPC systems may need substantial financial expenditures. Because of this, there are sometimes large lines of scientists and researchers waiting to use pricey, shared HPC workstations. For industrial and HPC applications, cloud computing provides new computing paradigms, capacity, and flexible solutions. Now, some of the computationally demanding applications that were previously run on conventional HPC machines might be run on the cloud. The cost model for cloud computing removes a substantial capital need. Thus, this research develops a two mechanism (PEAS and CEFT-HPC-Cloud) for efficient HPC-cloud model; The PEAS (Performance and Energy-aware Scheduling)-mechanism, which is intended for parallel computing with job scheduling and the best resource allocation in data centres, is introduced in this research study. After initially developing a system model for the parallel computing process, followed by the construction of a brand-new, efficient scheduling algorithm for job scheduling, an energy-aware mathematical model is built for the greatest possible use of energy. The assessment of PEAS takes into consideration Makespan, Energy consumption, and Power utilization as well as HPC-aware scientific operations like cybershake and montage workflow. Additionally, PEAS is more effective than any other model to this date. This research work proposes Multilevel CEFT-HPC-Cloud (Cost Effective Fault Tolerance) mechanism in HPC newline2 newlineCloud; CEFT-HPC-Cloud is a multilevel