近年来，低轨道巨型星座卫星网络(LEO mega-constellation satellite network，以下简称LEO 卫星网络) 因其广覆盖、低时延及高通量的通信潜力而受到来自工业界、学术界和政府机构的广泛关注。相比于传统地面网络，LEO 卫星网络所处空间环境更复杂，面临的威胁因素更多，卫星失效后果更严重，因此研究LEO 卫星网络容错技术是迫切且重要的。本文首先分析卫星失效对网络的影响及现有容错技术的不足，提炼出如下三个问题：（1）卫星失效下，如何保证网络流量无拥塞转发；（2）链路失效下，如何设计高可用性同时提高链路利用率的容错技术；（3）如何减少特定节点失效引发的拓扑剧变。针对这三方面问题，本文从全网流量规划和节点业务流调度角度，研究LEO 卫星网络的容错关键技术。论文的主要工作和创新点如下。(1) 提出基于流量工程的LEO 卫星网络主动容错机制。本文提出主动保护卫星节点和链路的容错技术SN-FFC。SN-FFC基于流量工程建立网络故障（节点和链路失效）容错模型，并基于线性规划和对偶理论降低模型计算复杂度。对网络中至多?个故障(𝑘为可选参数)，SN-FFC均能保证网络流量无拥塞转发。仿真结果表明，SN-FFC能够快速应对网络故障且性能开销较低。(2) 提出基于条件风险价值的LEO卫星网络容错机制。本文提出考虑不同链路故障概率的容错技术SN-CVaR。SN-CVaR设计基于条件风险价值的容错模型，以最小化条件风险价值为优化目标，建立概率故障模型，提出剪枝策略，使故障状态数减少5%-10%，降低模型计算复杂度。仿真结果表明，在相同可用性水平下，SN-CVaR能够支持最多两倍的流量。(3) 提出保护关键节点的LEO卫星网络容错机制。本文提出识别关键节点及保护策略的容错技术PKN。PKN建立基于累积时变图和复杂网络理论的节点关键度识别模型，提出最小化加权流组完成时间的保护策略PKN-WCCT。仿真结果表明，与随机删除相比，删除模型所识别的关键节点后，路径长度变长。所提保护策略能实现流组完成时间减少超过40%。
In recent years, LEO mega-constellation satellite network (hereinafter referred to as LEO satellite network) has attracted extensive attention from industry, academia and government agencies for its potential to provide wide coverage, low time delay and high throughput communication. Compared with the traditional ground network, LEO satellite network is in a more complex space environment, facing more threats, and the consequences are more serious when suffering satellite failures. Therefore, fault tolerance is urgent and important. This dissertation first analyzes the impact of satellite failure on the network and the shortcomings of existing fault tolerant technologies, then summarizes the following three problems : (1) how to ensure network traffic is forwarded without congestion in the case of satellite failure? (2) How to design fault tolerant technology with high availability and improving link utilization under link failure? (3) How to reduce topological upheaval caused by specific node failure. In view of these three aspects, thisdissertation studies the fault tolerant key technologies of LEO satellite network from theperspective of traffic planning and node flow scheduling. The main work and contributions of this dissertation are as follows.(1) An active fault tolerance mechanism for LEO satellite network based on traffic engineering is proposed. This dissertation presents a fault tolerant technique called SN-FFC for active protection of satellite nodes and links. SN-FFC models network failures (node and link failures) based on traffic engineering, and reduces the computational complexity of the model based on linear programming and duality theory. SN-FFC can guarantee the robustness of the network for up to ? faults (? is an optional parameter). Simulation results show that SN-FFC can quickly deal with network failures and has low performance overhead.(2) A fault tolerance mechanism for LEO satellite network based on Conditional Value at Risk(CVaR) is proposed. This dissertation presents a fault tolerant technique called SN-CVaR which takes into account different link failure probabilities. SN-CVaR designs a fault tolerant model based on conditional VaR, establishes a probabilistic fault model, and proposes a pruning strategy to reduce the number of fault states by 5% to 10% and reduce the computational complexity of the model with the optimization goal of minimizing conditional VaR. Simulation results show that SN-CVaRcan support up to twice the traffic at the same availability level.(3) A fault tolerance mechanism of LEO satellite network is proposed to protect key nodes. This dissertation presents a fault tolerant technology called PKN to identify key nodes and to provide protection. PKN establishes a node critical degree recognition model based on cumulative time varyinggraphs and complex network theory, and proposes a protection strategy called PKN-WCCT to minimize the completion time of weighted coflow. Simulation results show that compared with random deletion, the routinglength becomes longer after deleting the key nodes identified by the model.The proposed protection policy reduces coflow completion time by more than 40%.
