Renewable energy-aware demand response for distributed data centers in smart grid

Abstract

Data centers consume a large amount of energy from the power grid and have a high carbon footprint. Recent practice has introduced renewable energy to decarbonize the data-center operation. This paper studies how to leverage the demand response of a group of distributed data centers to exploit the diversity of renewable generations and energy prices in a smart grid. We model the data-center operation as an energy cost minimization problem, in which a data-center operator jointly optimizes the energy supply and computing workload allocation for distributed data centers. To solve this problem, we design a distributed demand response algorithm that involves little information exchange yet effective interactions between the operator and data centers. Specifically, the operator updates the dual variable and broadcasts to the data centers, and each data center optimizes its operation separately as a response to the dual variable. Using data-center workload traces and renewable energy data, we validate our solution method and the simulation results show the optimal computing workload allocation and energy management across distributed data centers.