Cloud computing requires huge data centers that consume significant amounts of electric energy. Such a significant energy consumer may be a risk factor for the power company since it increases the peak load and overload can occur, harming the equipment. This paper is motivated by the fact that the energy consumption of a data center is huge but elastic because the load in a data center can be transferred to other data centers. Since data centers pursue energy cost reduction, it is possible to use a data center heavily that is in a region where high incentives or low tariff rates are applied. To prevent overloads of power distribution networks, utilities need to determine incentives and tariff rates appropriately.
The major contribution of this paper is the solution of the energy consumption problem from the viewpoint of both data centers and utilities. This paper uses two-stage solutions: one is for data centers and the other is for utilities. To optimize the energy consumption for a data center, an objective function based on locational marginal price (LPM) and incentive-related parameters is used. The constraints include important characteristics for data centers, such as quality of service and energy cost per load. The objective function for utilities includes the overload risk, called electric load index (ELI), and cost for incentives. Simulation results show that the proposed method reduces ELI by 12 percent for utilities and energy cost by 46 percent for data centers, which provides a win-win solution to both sides.
According to the paper, the authors transform the optimization problems as linear problems and use a branch-and-bound algorithm. The authors exclude the details of the solution due to the page limit and refer to a technical paper. However, the link for the technical paper is not available through the Internet, which makes it hard to verify the solution.