Virtual machine placement quality estimation in cloud infrastructures using integer linear programming

Abstract

This paper is devoted to the quality estimation of virtual machine (VM) placement in cloud infrastructures, i.e., to choose the best hosts for a given set of VMs. We focus on test generation and monitoring techniques for comparing the placement result of a given implementation with an optimal solution with respect to given criteria. We show how Integer Linear Programming problems can be formulated and utilized for deriving test suites and optimal solutions to provide verdicts concerning the quality of VM placement implementations; the quality is calculated as the distance from an optimal placement for a given criterion (or a set of criteria). The presented approach is generic and showcased on resource utilization, energy consumption, and resource over-commitment cost. Experiments performed with different VM placement algorithms (including the VM placement algorithms implemented in widely used platforms, such as OpenStack) exhibit the competence of such algorithms with respect to different criteria.