Enhanced Evolutionary Computing Assisted Robust SLA-Centric Load Balancing System for Mega Cloud Data Centers

Abstract

Considering significance of a robust and Quality of Service (QoS) centric cloud computing, virtualization assisted load-balancing has been found a potential solution. However, assuring optimal Virtual-Machine (VM) migration with minimum violation of Service-Level-Agreement (SLA) and QoS degradation has been the challenge for academia-industries. VM allocation or scheduling being an NP-hard problem has been solved by numerous heuristic approaches such as classical Genetic Algorithm (GA), Ant Colony Optimization (ACO), etc. However they have been found confined due to local minima and convergence issues, especially for Mega Data Centres (MDCs). To alleviate such issues, in this paper an enhanced Evolutionary Computing algorithm named Adaptive Re-sampling GA (ARGA) algorithm has been developed that in conjunction with a stochastic prediction based dynamic load-measurement and Maximum Correlation (MC) assisted VM selection perform optimal load balancing over IaaS MDC infrastructures. The proposed ARGA VM allocation model with dual-level dynamic threshold assisted load estimation and MC based VM selection has exhibited lower SLA violation, performance degradation, downtime and minimum VM migration as compared to classical ACO based load balancing.