Optimal Virtual Machine Allocations in Cloud Computing System

Abstract

Cloud Computing (CC) is a popular computing paradigm by providing virtual resources as services over the internet. In the CC system, each user request can get services on a pay as you use basis. There is reward and system cost for serving each request, and the service provider can decide the amount of resources to assign to each request. Regardless of the random request arrivals and departures, a novel model of the Continous-Time Markov Decision Process (CTMDP) is explored. In this paper our goal is to maximize the total discounted expected reward for service providers. With the solution of the Bellman optimality equation by the value iteration method, an optimal policy can be achieved.