This Is AuburnElectronic Theses and Dissertations

Exploiting Renewable Energy and UPS Systems with a Renewable-Aware Scheduler to Boost Energy Efficiency of Data Centers




Peng, Xiaopu

Type of Degree

PhD Dissertation


Computer Science and Software Engineering


It is prudent to leverage on-site renewable sources like solar and wind to build environmentally friendly and energy-efficient data centers. Data centers deploy distributed UPS systems to handle the intermittent nature of renewable energy. As the pilot study of this dissertation research, we propose a renewable-energy manager, REDUX, to offer a smart way of managing server energy consumption powered by a distributed UPS system and renewable energy. REDUX maintains a desirable balance between renewable-energy utilization and data center performance. REDUX makes judicious use of UPS devices to allocate energy resources when renewable energy generation is low or fluctuate condition. REDUX not only guarantees the stable operation of daily workload but also curtails the energy cost of data centers by improving power resource utilization. Due to intermittent nature of renewable energy and fluctuating grid energy price, advanced data centers often deploy distributed UPS systems with high efficiency, scalability, and reliability. In the second part of this dissertation study, we investigate the problem of energy management for data centers with renewable resources and energy storage. As an extension project of REDUX system, we devise an interface-backbone layer framework designed unified energy efficiency management system called REDUX2. Properly allocating fluctuating renewable energy, UPS battery energy storage, and grid power with dynamic price, REDUX2 is primed to minimize the long-term electricity bill for data centers. With prediction of renewable energy supply, categorization of grid power price level, and energy storage in the UPS devices, REDUX2 orchestrates workload distribution facilitated by the heuristic algorithms that handle high-level control strategies such as renewable energy smoothing, UPS device control, back-filling, and non-urgent job admission deferrals. Compared with the existing solutions, REDUX2 demonstrates a prominent capability of mitigating average peak workload and boosting renewable-energy utilization. Leveraging on-site renewable sources like solar and wind provides ample opportunities on developing environmentally friendly and energy-efficient data centers. Evidence demonstrates that renewable-aware job schedulers conserve energy by adjusting the arrangement of non-urgent workload according to renewable energy states. To further upgrade REDUX2 to the next high level, we bring forth a unify designed energy management system - REDUX3 - equipped with a renewable-aware scheduler. REDUX3 judiciously schedules jobs submitted by users and manages the energy supply of data centers powered by grid, renewable energy, and distributed uninterruptible power supply (UPS) systems. REDUX3 seamlessly incorporates a renewable-aware workload scheduler to make back-fills and defer admission s of non-urgent jobs, thereby properly allocating fluctuating renewable energy, UPS battery energy storage, and grid power with dynamic price. Our experimental results confirm that REDUX3 is adroit at minimizing long-term electricity bills for data centers by boosting renewable energy utilization.