Placement of CPU-Intensive Virtual Machines in High Resource Utilization Cloud Datacenters. An Economical Revenue Maximization Analysis

Abstract

The enormous growth in the use of Cloud Service Providers (CSPs) leads to an increasing consideration of the optimization of Virtual Machine Placement (VMP) to host services for clients. This work aims to study VMP resolution algorithms in cloud datacenters with high resource utilization and CPU-intensive requested VMs for economical revenue maximization. Experiments were carried out in 64 different experimental scenarios. From the four evaluated algorithms in the experimental results, it can be seen that A1 offers the best results considering a centralized decision approach, First-Fit for the iVMP phase, Memetic Algorithm (MA) for the VMPr phase, prediction-based method for VMPr Triggering and update-based method for VMPr recovering. A1 slightly outperforms the other algorithms that also perform well for the analyzed scenarios considering average, maximum and minimum objective function evaluation metrics.