Fuzzy based energy efficient workload management system for flash crowd

Abstract

Many organizations are moving towards the cloud to meet sudden spikes due to a flash crowd. The exponential growth in requests and the offloading of the computations have increased data centers which in turn increases energy consumption and carbon footprint. This paper aims to propose an energy-efficient workload management scheme for the green cloud. The novelty of the work entails i) Developing a three-staged scheduling scheme that maps Virtual Machines with the active Physical Machines resulting in the consolidation of workload. ii) Deploying Fuzzy Decision Maker to analyze under-utilized and over-utilized servers based on the cores and the type of workload. The efficiency of the proposed system is tested through a trace-based simulation. The Power Usage Effectiveness benchmarked with standard algorithms claims that this work averagely conserves 1.9% energy by triggering 60% lesser migrations and reduces the execution time of cloud instances averagely by 6 clock cycles with a minimum Service Level Agreement Violations of 0.1%