Data Centre Optimization for Cloud Computing and Virtualization

Abstract

Cloud computing has significantly transformed modern data centre architectures by enabling scalable and flexible resource provisioning through virtualization. Despite its advantages, achieving energy efficiency while maintaining service reliability remains a critical challenge in virtualized cloud environments. Existing virtual machine (VM) placement and consolidation approaches often emphasize aggressive energy minimization, which can lead to excessive VM migrations and increased Service Level Agreement (SLA) violations under dynamic workloads.
This study proposes a Learning-Guided Adaptive Multi-Objective Virtual Machine Optimization (LG-AMOVM) framework that prioritizes stable, SLA-aware energy optimization in cloud data centres. The framework integrates workload trend prediction, adaptive utilization thresholds, and a net benefit-driven migration model to dynamically control VM placement and consolidation. Unlike static threshold-based or purely energy-centric methods, LG-AMOVM incorporates workload variability, migration overhead, total energy consumption, SLA compliance, and SLA risk into a unified optimization process, ensuring that migration decisions yield positive global benefits.
Experimental results demonstrate that the proposed approach consistently reduces energy consumption while maintaining zero SLA violations and limiting unnecessary VM migrations under dynamic conditions. Compared to baseline strategies, LG-AMOVM achieves balanced, adaptive optimization, demonstrating that energy-efficient cloud operation can be realized without compromising service stability.