Evaluation of different scheduling policies in IaaS applications by Mean Field analysis

Cloud Computing is emerging today as a commercial infrastructure that through the use of virtualization aims to provide on demand computing resources. In particular, the Infrastructure as a Service (IaaS) is a cloud service that allows the user to perform and execute any OS and application in virtual environments. In this work we present an evaluation of different policies for the assignment of virtual machines that IaaS providers can adopt in order to both efficiently planning their infrastructures and guaranteeing the quality of service expected by customers. The study is based on the exploitation of a Mean Field Model, able to evaluate systems with a relevant number of interacting entities, that can provide interesting insights derived from the computation of different performance indexes such as the request loss rate, the mean number of executing/available resources, and the overall ratio of satisfied requests to mention a few.