电力数据中心网络流控制理论及方法研究

Functioning as the central control unit of electric power system, a data center is the key support to integrated control power flow and information flow within smart grid. The heavy traffic in an electric power data center is prompted by power calculation systems. As a result, the network flows’ distribution and transmission have encountered numerous problems. Traditional traffic engineering (TE) is not able to solve these issues completely. This project will research electric power data center networks and respond to the new challenges：.(1).Investigating the characteristics of non-structured power data and the calculation of strong coupling relations among data, and proposing new splitting and distribution storage methods for data. This research will reduce the traffic load of data migration and ease the peak flows caused by migration in terms of data source aspects..(2).Investigating the double dimension controllability of data sources and data flows in data center networks, and proposing traffic-aware Virtual Machines (VMs) configuration methods and optimal VMs migration strategies. The research will balance traffic load in terms of traffic layout aspects..(3).Investigating transmission dynamic models of data center networks, determining this non-linear system’s stability with time-delay perturbations, and proposing novel traffic control methods. In the aspects of system control, the research will eliminate perturbations and stabilize self-oscillation, and the data center network’s performance will be enhanced..Electric micro-grid transient control experiments will evaluate the research practicality. The research outcome will also provide solid theoretical and technical basis for the development and operation of smart electric power system.

电力数据中心作为智能电网的中央控制单元，是实现电力流和信息流一体化控制的核心设施。然而中心内部网络流量受电力计算驱使，从流量分布到数据传输都呈现出新的特征，现有流量工程理论并不完全适用。本项目拟对电力数据中心网络流量的基础理论进行深入研究：研究电力非结构化数据特征和数据间的计算耦合性，提出数据拆分和分布式存储新方法，从数据源层面降低数据迁移负荷，平抑迁移时期的流量峰值；研究数据中心网络特有的数据源和数据流双维度可控性，提出流量感知的虚拟机配置新方法和最优迁移策略，从流量优化布局层面均衡网络负载；研究数据中心网络传输动力学模型，判定该非线性系统在时滞扰动下的稳定性和收敛性，提出新的流量控制方法和通信协议，从系统控制层面消除扰动和自激震荡，提升数据中心网络传输能力。并以微网小扰动暂态控制为实验用例进行实践验证，为推进我国电力数据中心的工程实用化进程提供坚实的理论基础和技术支持。

本课题在国家自然科学基金资助下，课题组深入研究了电力数据中心网络流量的基础理论问题：. 研究了电力非结构化数据特征，解析数据间计算耦合性，提出变长非结构化数据关联存储和数据源k重覆盖优化方法。从而有效降低数据迁移负荷，平抑迁移时期流量峰值。建立了电力系统分区求解的耦合计算任务数据交互关系的有向图模型，提出了任务改进关键路径算法和流量感知虚拟机配置新方法。从流量布局层面保证数据中心网络负载均衡。研究了受计算需求驱使的数据中心传输动力学模型，判定DCN非线性系统在单一小扰动和多维扰动矢量下的稳定性和收敛性，提出了基于数据流时间门限的改进虚拟机最优调度控制算法和协议，从系统层面平抑扰动和自激震荡，提升DCN传输能力。并以微网小扰动暂态控制为实验用例完成了数据中心网络时滞非线性系统扰动平抑过程实验，通过实验对理论算法进行了验证，实验研究为电力数据中心设计和构建提供丰富实践经验和技术支持。. 研究成果对推动我国电力数据中心科技发展和智能电网建设有着很好的科学意义和学术价值。

内容获取失败，请点击重试