Energy-aware virtual machine allocation and selection in cloud data centers

Data centers evolve constantly in size, complexity, and power consumption. Energy management in cloud data centers is a critical and challenging research issue. It becomes necessary to minimize the operational costs as well as environmental impact and to guarantee the service-level agreements for the services provided by the data centers. We propose a modified discrete particle swarm optimization based on the characteristic particle swarm optimization for the initial placement of virtual machines and a novel virtual machine selection algorithm for optimizing the current allocation based on memory utilization, bandwidth utilization, and size of the virtual machine. By means of simulations, we observe that the proposed method not only saves the energy significantly than the other approaches, but also minimizes the violations of service-level agreements.