针对大规模灾害的可靠通信网络设计与网络恢复

People now are increasingly relying on large-scale communication network infrastructures (like optical backbone communication networks) to provide reliable and mission critical communications for disaster rescue, datacenters, government, military, finance, health care, etc. It is notable, however, that such communication network infrastructures themselves are facing more and more potential disaster threats, both natural, like as earthquakes, hurricanes, and man-made, like terrorist attacks. Some recent major network disruptions due to disasters include 2012 Sandy Hurricane, 2011 Japan Tsunami, etc. Despite a lot of research activities on reliable network design, a complete theory for disaster-resilient network design has remained largely unexplored. By now, the lack of such a general theory has significantly hindered the systemic design of future highly disaster-resilient network infrastructures with the aim of promoting a safer life for human beings. The goal of this project is to develop a complete theory for disaster-resilient network design by investigating its three fundamental issues, namely disaster modelling and related network reliability assessment, proactive network design and protection against disaster, and reactive pre/post-disaster restoration. A number of important topics, including the modelling of disasters with the considerations of both spatial failure correlation and temporal failure correlation, cost efficient assessment of network reliability against disaster, disaster-aware network design, disaster-aware resource provisioning, virtual router-based network protection, pre-disaster restoration based on disaster alerts, failure localization under disaster, post-disaster restoration through both immediate relief and intelligent relief techniques, will be fully explored in this project for receiving a wide understanding on the disaster-resilient network design. We envision that this project will solidly contribute to the development of modern Network Theory and to the continuing success of future communication networks.

作为国家重要基础设施的通信网络在面对自然灾害等大规模破坏时展现出脆弱性的一面，而随着网络业务的爆炸式增长和人们对网络的依赖加强，亟需扩展目前的网络可靠性理论来应对灾害对网络的影响。本项目的目标是开发一套完整的具有容灾能力的网络设计理论。该工作将集中研究容灾网络可靠性理论的三个最根本问题：1）灾害建模和相关的网络可靠性评估；2）主动的网络防灾设计和网络保护；3）灾前与灾后的网络应急恢复。许多重要的课题将在这个项目中得到充分探讨：考虑空间故障相关性和时间失效相关性的灾害模型设计；高效的防灾网络可靠性评估；灾害敏感网络设计和网络资源配置方案；基于软件定义网络的故障影响最小化策略；根据灾害警报信息的前期预防措施；针对灾害区域特点的故障定位；紧急和智能的故障缓解措施以及灾后恢复设计。通过对区域故障这一类问题展开更深入的研究，我们期望该项目将扎实地为现代网络理论的发展和未来通信网络的持续成功做出贡献。

互联网技术的迅速发展以及大数据时代的到来，人们对网络的可靠性要求也越来越高。然而，频发的灾难性事件因其破坏巨大、影响持续时间长，成为通信网络的主要威胁之一。在面对位置和规模都不确定的区域性故障下，如何评估目前已有网络的可靠性，设计高效快速的网络保护和网络恢复技术是本项目关注的重点。近年来，几何概率分析技术的发展，为具有空间故障相关性的灾害影响分析提供了必要的理论支持。我们利用该理论，设计了基于采样算法的快速收敛的网络脆弱性分析工具。软件定义网络技术将传统网络的数据平面和控制平面分离，使得全网灵活重配置和管理成为可能。我们通过解决集中式控制器的可扩展性、包的一致性等问题，提出了基于软件定义网络技术的低开销网络保护策略，并实现了系统原型。进一步，我们探索利用网络虚拟化技术实现大面积多故障下的网络故障快速恢复策略。我们利用共有云数据中心提供的系统可靠性基础上，采用overlay routing技术，通过多跳、多路径路由实现快速重路由故障恢复，保障在大规模区域故障下的通信生命线。此外，在该项目支持下我们还开发了一系列移动计算应用，提升用户安全和认证机制。我们针对公共交通工具震动程度的分析工具可以对公共交通系统进行评估；针对行人安全保护工具，保证用户在安全区域行走，避免到达危险区域；此外我们还设计了用户认证工具，可以通过用户的自定义声纹密码，在嘈杂或无声情况下对用户进行认证。在高性能网络交换理论方面，设计了目前复杂度最优的光优先级缓存结构。关于网络脆弱性分析和容灾保护的工作被加州大学戴维斯分校B.Mukherjee教授(IEEE Fellow)在多篇论文中作为代表性工作与 MIT 研究组、哥伦比亚大学研究组相关成果进行比较分析；所提出的网络损毁度概念、模型及测定理论受到国际同行关注，应邀在欧洲光通信学术会议 NOC’14 做特邀报告，在华为主办的云可靠性工业界国际论坛 CRW’16 做 Panel Speaker 报告。在该项目支持下共完成论文18篇，申请10项专利，并开源多个可靠性网络支撑平台。