Automated capacity management and selection of infrastructure-as-a-service providers

Infrastructure-as-a-service (IaaS) providers are gaining popularity in the application hosting market. Virtual machines of different sizes (capacities for server resources such as CPU, main memory, etc.) are offered on-demand on an hourly, daily, weekly or monthly basis. Which offering minimizes the hosting costs depends on the sizes and respective prices, but also on the demand patterns of an application and is a nontrivial selection problem for the customer. The fact that IaaS providers have different minimum subscription times, makes the problem particularly hard. In this paper, we describe an optimization formulation that determines an optimal schedule of virtual machines, which is needed to satisfy the demand of a particular application. The algorithm can be used to select the cost minimal offering of different hosting providers, but also to control the allocation and de-allocation of virtual machines with an IaaS provider over time.