Towards Economic Energy Trading in Cloud Environments

Especially in times of heavy loads, cloud providers often have to outsource tasks to external clouds to fulfill service level agreements. Nevertheless, a cloud provider maximizes the company's benefit while running as many jobs as possible on the own hardware without going below a specific workload of the running processors. Since cloud providers will have to estimate the required energy in advance due to energy trading, they should aim for estimating maturely the optimal number of necessary processors for a future date and time. This paper presents a method for anticipating the optimal number of active processors and corresponding energy. In particular, this work analyzes the potential of Gaussian processes to estimate future jobs by considering statistical data. Based on the job number estimate, a second Gaussian process approximates the optimal number of processors for a future date allowing for economical energy trading. Finally, the paper optimizes the computing resources in clouds by applying earliest deadline first strategy.