Energy-based accounting and scheduling of virtual machines in a cloud system

Abstract

Currently, cloud computing systems using virtual machines bill users for the amount of their allocated processor time, or the number of their virtual machine instances. However, accounting without cooling and energy cost is not sufficient because the cooling and energy cost is expected to exceed the cost for purchasing the servers eventually. This paper suggests a model to estimate the energy consumption of each virtual machine. Our model estimates the energy consumption of a virtual machine based on the in-processor events generated by the virtual machine. Based on the suggested estimation model, this paper also proposes a virtual machine scheduling algorithm that conforms to the energy budget of each virtual machine. Our evaluation shows the suggested schemes estimate and provide energy consumption with errors less than 5% of the total energy consumption.