面向绿色数据中心的高效能分布式储能技术

信息时代的能耗和环境问题使得绿色数据中心备受瞩目。分布式储能最近成为绿色数据中心中新兴的重要技术: 它能削减负载峰值功耗带来的运营成本，还能便于融合新能源来限制碳排放。然而以往研究忽视了储能设备本身的结构优化，因此面临着一个关键设计瓶颈。目前方案多局限于利用单一同构的储能技术完成简单的电力缓存任务，它们非但缺乏应对特殊功耗波动的能力，还难以保证系统的额定使用寿命。更重要的是，若没有一种跨层次的管理，数据中心会因缺失备份能源而面临宕机风险。本项目将探索新型的数据中心储能方法：一种跨越储能设备和软硬件的综合电源管理技术。研究将侧重三个关键科学问题，即， 如何设计混合异构储能系统来管理复杂供电异常，如何均衡使用储能设备从而提升其使用寿命，及如何协同调度负载以降低供电失效风险和优化系统全局可用性。本项目将有效提升数据中心的可持续性和经济性，并为新型绿色中心提供有力的支持。

Green data center design is attracting growing attention due to the heightening demand for alleviating the increasing IT energy needs and resolving the looming environmental problem. Recently, distributed energy storage (DES) systems have emerged as the key enablers that allow a green data center to reduce operating expenditure by shaving power spikes and to cap carbon emission by incorporating renewable energy sources. Nevertheless, these important power systems could become the major efficiency bottleneck in future green data centers. Prior studies are primarily limited to using battery-based DES technology and homogeneous DES system as a simple energy buffer, while neglecting the architectural optimization of DES. Therefore, existing designs not only lack the capability of efficiently handling irregular server power spikes, but also fail to maintain the desired service life of their DES systems. More importantly, without a cross-layer power management, current proposals can frequently deplete the critical backup energy and pose serious threat to service availability. This project aims to investigate a new data center DES design and management approach, which exploits the benefits of hybrid and heterogeneous DES. Specifically, we propose to design a holistic power management framework that spans workload application, server hardware, and energy storage. We will focus on three critical questions: 1) how to design hybrid and heterogeneous DES to manage various power anomalies; 2) how to balance the usage of DES units to improve their lifetime and cost-effectiveness; 3) how to cooperatively manage workload application to reduce the risk of power failure and improve the overall system availability. The proposed research can greatly improve the sustainability and economy of data centers, and will pave the way for a new class of green data center design.