Energy-aware VM Placement with Periodical Dynamic Demands in Cloud Datacenters

In cloud datacenters, energy-efficient Virtual Machine Placement (VMP) mechanism is needed to maximize energy efficiency. Existing virtual machine (VM) allocation strategies based on whether the VMs' resource demands are assumed to be static or dynamic. Apparently, the former fails to fully utilize resources while the latter, which is implemented on shorter timescales, is either complicated or inefficient. Moreover, most prior VMP algorithms place VMs one by one, which lacks an optimization space. To handle these problems, we predict Gaussian distribution patterns of VM demands and propose an ant-colony-system VM placement algorithm (GACO-VMP) which synchronously coordinates the VMs with complementary resource requirements on the same server. The Gaussian distribution pattern is derived from the VMs running the same job. This mechanism minimizes energy consumption, while guaranteeing high resource utilization and also balancing resource utilization across multiple resources. In addition, we design two new metrics, called cumulative utilization ratio(CUR) and resource balance distance (RBD), in order to measure the overall resource utilization level and the equilibrium of multi-dimensional resource utilization, respectively. Simulations based on Google Cluster real trace indicate that GACO-VMP can achieve remarkable performance gains over two existing strategies in energy efficiency,VM migrations, resource utilization and resource balance.