A cyber-physical systems approach to energy management in data centers

Abstract

This paper presents a new control strategy for data centers that aims to optimize the trade-off between maximizing the payoff from the provided quality of computational services and minimizing energy costs for computation and cooling. The data center is modeled as two interacting dynamic networks: a computational (cyber) network representing the distribution and flow of computational tasks, and a thermal (physical) network characterizing the distribution and flow of thermal energy. To make the problem tractable, the control architecture is decomposed hierarchically according to time-scales in the thermal and computational network dynamics, and spatially, reflecting weak coupling between zones in the data center. Simulation results demonstrate the effectiveness of the proposed coordinated control strategy relative to traditional approaches in which the cyber and physical resources are controlled independently.