【作者】 李彬；

【导师】 陈向东；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2007， 硕士

【摘要】 多协议标签交换(MPLS)是IP骨干网中前景良好的关键技术之一。在MPLS网络中，数据包通过定长的标签作为索引进行转发，较之于传统的IP路由，它对业务流量有着更灵活的控制与管理。作为骨干网交换技术，MPLS需要对故障进行更快速的响应，否则可能导致业务QoS下降。IETF将MPLS的故障恢复模型划分为两大类，重路由和保护切换(快速重路由)。二者的差异在于前者是通过在线建立新路由以传输流量，而后者刚需要预先建立保护路径以切换流量。本文将针对这两种模型分别进行了研究，主要工作如下：其一，在区分服务网络环境中，描述了一种基于集中化模型的多故障的重路由算法。采用以优先级属性划分LSP的策略，将带宽作为主要需求参数，满足了不同LSP的QoS要求，其实现是基于MPLS流量工程的约束路由的最优化方程。故障后，高优先级LSP优先占有网络资源，在网络资源不足时，高优先级LSP通过对低级别流量实施抢占获得新路由，确保了重要业务流的QoS。模拟示例分析验证了算法的可行性和优越性。其二，在备份路径预建立的前提下，提出了一种MPLS的故障恢复方案——EB。该方案在现有的Bridge模型基础上引入了修改后的Backpressure信令以解决分组失序。此外，结合了现有了经典方案(Haskin，Makam)，进行了理论性能评估及仿真实验分析。总体表明，EB良好继承了Bridge方案原有的优势，并且解决了分组失序问题。综合评估，EB较之现有的几种方案有着更优恢复性能。其三，提出了一种基于QoS保护的MPLS的故障恢复方案P(Proposal)。该方案结合了段保护和路径保护的思想，在段保护域中引入了现有的Haskin和Makam的混合思想，实施局部回传以减小Haskin和Makam的恢复时间。与此同时，也减小了Local Reverse方案的平均时延。此外，文中扩展了现有的故障恢复的QoSP评估模型，并对建议方案进行了评估分析。实验证明，针对区分服务的四种流量类别，P方案均有着具有较优的QoSP性能。更多还原

【Abstract】 Multiprotocol Label Switching (MPLS) is the key technology of backbone IP network, with good prospect. In MPLS network, forwarding of IP packet is driven by the fixed-length labels. Compared with conventional IP forwarding, MPLS has more flexible control and management for the traffic service. As the key switch technology of backbone network, MPLS has to response as fast as possible, otherwise, it will result in degradation of QoS level.IETF(Internet Engineering Task force)divides the MPLS-based recovery into two models:. reroute and protection switching(fast rerouting). Recovery by rerouting model is defined as establishing new paths on demand for restoring traffic after the occurrence of a fault. But protection switching mechanism need pre-establish a recovery path or path segment to switch working traffic. That is the just chief difference between the two models. The thesis expatiates the research aiming at the two models respectively, and the central achievements is shown as follows:Under the Diff-Serv environment, this thesis described an algorithm for rerouting which is based on Multi-failures model and centralized implementation model. Dividing the LSPs according to the set-up priority and holding priority of LSP, the algorithm focuses on the bandwidth requirements for each LSPs in order to meet the different QoS requirements of LSPs. Furthermore, it is carried out by a optimization equation based constraint route of MPLS TE. after the occurrence of failure, LSPs with high-priority will possess the network resource with precedence, as well as preempt the LSPs with low-priority while there is lack of network resource .It can ensure the QoS level of the critical service traffic. At last, the case study has validated the feasibility and the advantage.In the case of that backup path has been pre-established, a scheme named EB for MPLS-based recovery is present, which introduces the modified Backpressure signaling mechanism to Bridge model in order to eliminate the packet disordering. In addition, combining the two existed classical scheme (Haskin, Makam), the thesis launch both theoretical analyses and simulation experiment to prove out the recovery performance. General indicating, EB not only preserves advantages of Bridge but also solves the packet disordering. Comprehensively evaluating, comparing with existed scheme, better performance is shown by EB.Besides, the thesis proposes an enhanced scheme based QoS protection, which adopts a hybrid approach of segment protection and path protection. In detail, Introducing the combination way of Haskin and Makam, locally returning in SPD(Segment Protection Domain)is implemented in order to minimize the restoration time of Haskin and Makam, as well as the average delay of existed scheme Local Reverse. Meanwhile, the evaluating function of QoS protection for MPLS recovery is enhanced to evaluate Haskin, Makam, Local Reverse, and proposed model. The experiment results indicate that better performance is shown by the proposal. The experiment results that better performance is kept up by the proposal, for the four different traffic classes under Diff-Serv.更多还原