Objective: This project aims to improve the energy efficiency of data centers significantly by introducing coordinated control of the cooling systems and the information technology. This requires new integrated, control-oriented models of data center dynamics representing thermal interactions between the computational system and the cooling system as a function of the control inputs to each system. New metrics will also be developed to capture effectively the impact of control decisions on computational performance in the context of service level agreements (SLAs) and other objectives and constraints that arise in practice. Multi-level, distributed control strategies will be developed to deal with the wide range of time-scales and spatial distributions that characterize data center dynamics. Intellectual Merit: This transformative research will lead to the first data center control strategies that explicitly coordinate cooling system actions with workload and power state allocations to achieve effective tradeoffs between computational performance and energy consumption. These strategies will also leverage new degrees of freedom by relaxing performance requirements as allowed by SLAs. The control hierarchy is based on abstractions that reflect the range of spatial and temporal scales common in many large-scale cyber-physical systems. Broader Impact: The academic-industry collaboration in this GOALI project assures the results will be implementable in future data centers. This will lead to significant reductions in the energy required to meet the demands for computation and storage in the U.S. for years to come. Through courses and internships, engineering students will be introduced emerging concepts in cyber-physical systems research.
