A cost-efficient mechanism for dynamic VM provisioning in cloud computing

Cloud computing gradually becomes an issue for modern IT business because of its flexibility, convenience, and low cost. It provides resources with guaranteeing QoS(Quality of Service) to cloud users and revenues to service providers. Due to the dynamics of user demands, service providers need VM(Virtual Machine) provisioning mechanism to predict the amount of resources demanded by cloud users and to estimate the amount of the resources prepared by service providers in the near future. In this paper, a new VM provisioning mechanism is proposed to elastically estimate the amount of resource to be prepared. We present an optimization model of the estimation by maximizing a provider's revenue while satisfying users' QoS requirements and minimizing the penalty for SLA violation. To evaluate the effectiveness of our solution, we consider many practical situations of a cloud system with a series of real trace data. It shows that our model elastically adapts to the dynamic environment while maximizing the revenue of a service provider.