【作者】 何涛；

【导师】 王锁萍；

【作者基本信息】 南京邮电大学 ， 信息网络， 2011， 博士

【摘要】 QoS路由是网络QoS控制技术的一个关键技术,它的基本任务是为一次连接寻找一条有足够资源、能满足QoS要求的可行路径。随着各种业务量的增加,网络拥塞现象日益严重,负载均衡的目的就是实现网络资源的优化使用,提高网络资源的利用率。基于负载均衡的路由算法可以使网络中的不同链路、路由器和交换机之间平衡业务负荷,更好地避免部分资源被过度使用而另外一部分为未充分利用。随着无线宽带网络技术的快速发展,频谱资源愈显“匮乏”,能够提供“伺机接入”的认知无线电技术开始被重视,这也给路由算法带来了新的挑战。论文主要研究了不同网络模型下的QoS路由与负载均衡路由算法及相关技术。时延约束下的最小费用路径问题是一个典型的QoS路由问题,论文提出了一种基于拉格朗日松弛法的路由算法,通过将链路费用参数吸收到延时参数中,同时在迭代过程中结合延时约束条件,可以在多项式时间内找到一个最优或较优解,并对该算法进行了分析和仿真,证明了算法的正确性。蚁群算法属于仿生学算法,被大量应用于求解NP完全问题。论文针对实际网络的特点,提出了并行迭代方式的蚁群优化算法,并通过仿真说明了该算法在性能上与采用串行迭代方式的算法持平但在实际消耗的时间上远小于串行方式。如何根据带宽请求合理建立LSP,提高网络资源的利用率,是MPLS的重要研究方向。论文在分析最大流算法的基础上,提出了MCRA算法,在选择路径时,算法同时考虑到了最大流零流边对最大流的影响以及MPLS网络中各出口-入口对之间带宽资源的竞争。仿真结果表明,与参与测试的算法相比,最小竞争路由算法能够接受更多的路由请求。针对支持多速率传输的无线Mesh网络,以博弈论中的循环囚徒困境的“针锋相对”战略为启发,论文提出了节点间亲密度的概念,并以此为因子激励节点间的合作。节点是否对RREQ分组进行转发取决于是否符合自身利益,它们的决策影响到相关节点间的亲密度。最后通过仿真证明了该方法能够有效地改善网络性能。在认知无线电Mesh网络中,由于此用户采用“伺机接入”方式与主用户共享信道,因此给定数据传输需求和有限频谱资源限制的条件下,为了能够在最短时间内完成用户的数据传输需求,必须将路由、调度和频谱分配联合考虑。频谱分配算法是认知无线电网络中路由算法的关键技术之一,为此提出了基于图论的分配算法,算法的目的有两个,一个是最大化分配的信道带宽,另一个是最大化同时活动的用户数。算法结合了染色模型和二分图匹配模型的优点,仿真结果表明算法改善了CR网络的性能。更多还原

【Abstract】 QoS Routing is an important technology of QoS control system. To find a feasible path that has sufficient resource and can meet the requirements of QoS constraints is a basic task of QoS Routing. With the rapid growth of various traffics, especially the multimedia traffic, the congestion of internet becomes more and more serious. Load balancing is one of solutions that proposed to solve the congestion problem. Load balancing aims to optimize resource of network and improve the efficiency of utilization of resource. Load balancing routing algorithms make the traffic load of links, routers and switches in network smoother. It can avoid that a part of resource of network be overused while others not be used efficiently. With the development of technology of wireless networks, spectrums become more and more lack. Therefore, cognitive radio (CR), a wireless network with opportunity access, aroused interesting of many researchers. But it brings a different difficult problem to the routing algorithm research. This paper focus on QoS Route, load balancing route and relative technology.DCLC (Delay Constrainted Least Cost) problem is a famouse problem of QoS Route. A novel approach based on Lagrangian relaxation was proposed. In this approach, the cost parameter of links is absorbed into delay parameter and taked the delay parameter into account whiles each iteration process. This approach can find a feasible path during non-polynomial. Simluations and analysis proved the correctness of porposed algorithm.Ant Colony Algorithms are one of bionics algorithms and are used for solving the non-polynomial completed problems usually. According to the characteristic of network, a new ant colony algorithm was proposed. The new algorithm adopts parallel iteration mode while a new routing request arrival. Simulations show that the proposed algorithm with parallel iteration mode has the same performance with the algorithms adopted the serial iteration mode but spend less time.It is very important for MPLS networks to build a LSP (Label Switched Path) reasonably. Bandwidth is a key constrained parameter of LSP request usually. How to choose a reasonable path which can both satisfy the request of bandwidth and improve the utilization of resource of MPLS networks is a difficult problem. Based on the maximal flow algorithm, a novel algorithm, named MCRA, was proposed. While finding path, MCRA considers not only the influence of maximal-flow-zero-edge, but also the competition of bandwidth between every ingress-egress pairs. Simulations show that, comparing with other algorithms, MCRA can accept more LSP request for it can balancing the load more efficiently.Recently, more and more wireless networks begin to support multi-rate transmission. It makes the routing problem of wireless mesh network (WMN) more and more complex. In this paper, a novel routing algorithm for multi-rate wireless mesh network is presented. It aims to achieve load balance and support multi-rate effectively. Alternating prisoner’s dilemma is a famous problem in game theory. Tit for tat (TFT) strategy is the best solution of this dilemma. Inspired with TFT, a new conception, named consanguinity, is been proposed by this new algorithm. This conception is used to inspirit nodes of network to cooperate with each other. The users with higher consanguinity will glad to forward RREQ packet. Simulation result shows that the proposed approach can improve the performance of WMN effectively.In CR mesh network, the secondary users (SUs) share the spectrum with primary users (PUs) by opportunity access mode. The SUs must quit the using spectrum as soon as PUs appear in this spectrum. It means that must transmit its data as soon as possible. So, in CR mesh network, routing algorithms should take the spectrum allocation problem into account. The spectrum allocation algorithms are one of important domain of routing algorithms research in CR mesh network. This paper proposed a new approach, which is based on graph-theory and aims at two objectives: one is the sum of the allocated channel bandwidth is maximum, and the other is the number of users can be active simultaneity is maximum. The simulations show that the presented algorithm can improve the performance of spectrum allocation of CR mesh network.更多还原