Energy Efficient Scheduling Based on Marginal Cost and Task Grouping in Data Centers

High energy consumption of data centers has been a widely concerned issue and has become a research hotspot. Recent studies show that both servers and cooling systems are main energy consuming factors of data centers. However, most of these studies only consider energy optimization either in servers or the cooling system, separately. Thus, it is essential to develop a scheme to jointly optimize the energy consumption of servers and cooling systems in the data center. In this paper, a joint energy optimization scheme is proposed to coordinately optimize the energy consumption of servers and the cooling system. Firstly, in the cooling system, different cooling modes, including the outside air cooling and liquid cooling are jointly considered. An energy consumption model for the cooling system is designed, where a segmented function is used to describe different energy consumption rate at different stages of the cooling system. A strategy is designed to use different proportion of the outside air cooling and liquid cooling according to real-time workload characteristics. Then, combining the energy consumption of the cooling system and servers, a joint energy optimization problem for data centers is formulated. Meanwhile, an energy-efficient task scheduling strategy based on marginal cost and task grouping is developed for solving the problem. Simulations have been conducted based on real-world workload traces, and simulation results demonstrate that the proposed approach outperforms previous techniques in energy savings.