Energy-Aware Virtual Machine Placement in Cloud Data Centers Using an Enhanced Multi-Objective Grey Wolf Optimizer

Abstract

Server consolidation and power consumption in data centers become more complex as server loads increase and virtual machine placement (VMP) becomes a key issue. This paper proposes a static VMP method for heterogeneous cloud data centers using the Enhanced Energy-Aware Multi-Objective Grey Wolf Optimizer (EA-MOGWO). It is implemented through energy-aware seeding, alpha-beta-delta-based discrete host representation with leader guidance, adaptive leader inheritance and mutation management, feasibility repair, fixed-reference objective-based normalization, and bounded local consolidation. The fitness function considers normalized server power, the number of active hosts, the migration ratio, and CPU-RAM imbalance, with weights of 0.40, 0.25, 0.20, and 0.15, respectively. We conducted experiments on a custom-made Python simulator with 50 heterogeneous hosts, 100, 200, and 300 VMs, a population size of 15, and 10 independent paired runs with 30 iterations. In all scenarios, EA-MOGWO had the lowest mean weighted fitness. The mean fitness decreased by 22.0%, 11.9%, and 3.6% relative to the PSO baseline, and by 28.8%, 25.8%, and 22.1% relative to the GWO baseline, respectively. The relative power reduction was greatest at 100 VMs and gradually closed as the workload density increased. The results mainly reflect better consolidation from increased migration and CPU-RAM imbalance in lighter workloads. The findings confirm EA-MOGWO as a static VM-placement approach for consolidation purposes in the considered CPU-RAM model and encourage its extension to larger workloads, different SLA parameters, and production traces.