Energy Aware Resource Efficient-(EARE) Server Consolidation Framework for Cloud Datacenter

Abstract

Cloud datacenter offers economic and elastic computing benefits to the customers, where on-demand virtual machine (VM) allocation plays a significant role. Inefficient VM placement leads to resource wastage and high power consumption that raises the requirement of server consolidation. The feasibly optimal placement of VMs with the objectives of minimum power consumption and maximum resource utilization is the key to server consolidation. Though many multi-objective VM placement schemes are available in the literature, it mostly works on weighted sum approaches that transforms multi-objective problem (where some objectives maximize and others minimize) into single objective to give optimized solution. Hence, the existing approaches do not correctly justify the multi-objective VM placement problem. To provide correct solution of multi-objective and multi-constrained VM allocation problem, this work presents GA based evolutionary server consolidation framework by applying rank based non-dominated sorting for multiple objectives to generate pareto-optimal solution. It enables maximum resource utilization and minimum power consumption to accomplish effective server consolidation. The VM placement is done using genetic algorithm (GA) which encodes VM allocation information into chromosomes. The performance evaluation of the proposed work is carried out by execution of numerous experiments in simulated datacenter environment. The experimental outcome reveals that the proposed VM allocation framework improves resource utilization upto 38.54, 41.67, and 44.8% and minimize power consumption upto 11.32, 12, and 13.7% over random, best-fit, and first-fit heuristic-based approaches.