【作者】 朴红艳；

【导师】 杨素英；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2008， 硕士

【摘要】 传统IP网络的路由体系只能提供数据传输的可达性服务，有的网络资源经常出现有些链路被过度使用，有些链路却空闲不用的负载不均衡局面，致使网络拥塞，业务的服务质量QoS(Quality of Service)也无法得到保证，这说明当前因特网没有动态的路由调整和灵活的网络控制能力。目前，多协议标记交换技术MPLS(MultiProtocol Label Switching)被公认为是实现流量工程TE(Traffic Engineering)的最有力工具，如何通过MPLS技术找到能够更好的解决网络路由中优化控制问题成为当今的研究热点。论文在深入研究MPLS网络中实现流量工程TE的基础上，着重于优化路由问题，针对传统路由算法存在的局限性，提出了两种路由算法——多约束路由选择的改进遗传算法和动态多路径负载均衡LDM改进算法。多约束路由选择的改进遗传算法借助遗传算法适配函数可实现多目标约束的优势，目标是在满足带宽、时延的基础上，将资源消耗和负载均衡分布作为优化目标，希望在消耗较少网络资源的同时，使负载尽量分布在宽裕或空闲资源的链路上，便于今后接纳更多请求。动态多路径负载均衡LDM改进算法根据MPLS支持多路路由特点与现有负载均衡算法单路路由的不足，针对路径负载率和最小跳数两个QoS属性而提出的。通过构造可能性系数作为选路标准，用较好候选路径集代替单一最佳候选路径传输流量。在流量分配中引入与可能性系数相关的负载比例系数，有利于降低计算的复杂度。论文针对提出的两种算法分别进行仿真验证，仿真结果表明多约束路由选择的遗传算法能够提高网络吞吐量，使负载分布更趋于合理，改善网络的传输服务质量，而改进LDM算法在流量不断增加状态下仍可有效避免网络拥塞，使网络负载尽量分布均匀达到合理分配资源的目的，同时对网络路径总延时的减小也得到明显改善。课题的研究从动态的优化网络路由控制和网络负载的均衡调节的两个方面，对目前网络流量的负载不均衡而导致网络瓶颈效应局面得到进一步改善，具有现实意义上的应用前景。这样的流量工程TE路由方式将更多还原

【Abstract】 Traditional IP Routing system can provide accessibility of data transmission services, which always leads to the unbalanced traffic distribution, some links getting over-utilized, while others remaining under-utilized, which results in congestion and the quality of services not guaranteed. That shows the present IP network lacks of the means to manage network resource efficiently and cannot make full resource utilization. At present, MPLS has been regarded as a powerful tool to traffic engineering. How to resolve the question about the optimal control of network routing by MPLS is always a hot point.Basing on survey the methods of realizing the traffic engineering, emphasizing the optimal routing, aiming at the limitations of traditional methods, putting forward two routing algorithms, an improved GA of routing multi-limited algorithm and an improved dynamic multipath LDM algorithm. The improved GA of routing multi-limited algorithm is in virtue of the advantages of GA, satisfies the needs of bandwidth and delay, makes the source consume and distribution load balancing as the goal, hopes consuming less network resource distribute the traffic into under-utilized path, accepts more requests. The improved dynamic multipath LDM algorithm uses the multipath character of MPLS and the shortages of single path routing, aims at two QoS attribution including load rate and the least hop. Using the possibility as choosing path’s standard, the traffic is proportioned among a few "good" paths instead of routing it all along the "best" path. The proportion algorithm can reduce the complexity of calculation.The experiments show the improved GA of routing multi-limited algorithm can improve the network throughput, reasonable distribution load balancing, enhance the quality of transmission services, while the improved LDM algorithm also can avoid congestion when the traffic increased rapidly, gain resource utilization efficiently, at the same time, the network delay reduced evidently. From both dynamic optimization of network routing control and the balance network load regulation, the two algorithms researched in this dissertation can further improve network bottleneck effect situation, which have the realization foreground of application.更多还原