Model predictive control for the placement of virtual machines in cloud computing applications
Mauro Gaggero, Luca Caviglione · 2016
Placement is the process of deploying virtual machines (VMs) over the physical machines (PMs) available in a cloud datacenter. Unfortunately, too many running PMs inflate energy requirements, while too aggressive packings of VMs over the same host degrade performances. Therefore, the paper presents a VM placement method based on model predictive control to reduce the power consumption of cloud datacenters while maintaining Quality of Service requirements. To describe the evolution of the system, a discrete-time dynamic model is introduced with several constraints. Placement strategies are obtained by solving finite-horizon optimal control problems with integer variables at each time step. The effectiveness of the proposed approach is evaluated through simulations and compared with two heuristics taken from the literature.