DS-TE应用MPLS VPN网络解决时延问题的研究

发布时间：2018-12-12 13:09

【摘要】：多协议标签交换核心之处就是标签概念的引入和应用。多协议标签交换主要有CE、PE、P三部分组成，CE也叫用户边缘路由，主要为用户接入VPN业务，PE也叫服务提供商边缘路由，它与CE直接相连，具有MPLS网络中LER的功能，同时还为每个用户创建一个虚拟路由转发表，以此实现多个VPN用户的隔离和VPN业务标签的分发。P也叫骨干网核心路由，它具有MPLS网络中LSR的功能，完成快速转发和控制任务。 DS-TE技术是区分服务模型和流量工程模型的集合，在区分服务模型中数据分组根据业务等级分为不同的优先级，体现在封装中的TOS字段，由6位二进制数组成，共分为12个级别，最高级别属于网络控制类；最低级别属于尽可能服务类。当MPLSVPN接收到具有区分服务等级的数据包时PE(LER)会将PHB映射成标签中的EXP字段，将12个优先级转化成8个优先级，然后根据不同的优先级选择不同LSP资源。流量工程模型在MPLS VPN中的作用是防止过多同类型或同优先级的转发等价类占用同一个或同几个LSP资源而造成其他一些带宽资源被浪费，MPLS TE所需要使用的功能部件大致由四个部分组成：分组转发、信息分发、路径计算和信令部件。所以，流量工程很大程度上可以减少VPN网络中的拥塞的情况发生。因此，DS-TE技术是一种既满足了用户的通信需求又能为运营商避免资源浪费情况的双重保障技术。 BH算法出发点是解决带宽浪费的问题，所以算法的抢占原则和代价公式都是以带宽的利用率为主。当前的计算机通信网络由于光线通信的发展已经拥有非常丰富的带宽资源，国内的各大运营商也陆续开始对带宽业务降价促销。所以企业用户的可用的带宽资源已经不再是限制通信服务质量的瓶颈。BH-PREPT算法在避免级联抢占方面有较好的性能，缺点是过多的关注了由于抢占LSP造成带宽浪费的代价而使算法的复杂度和对设备要求提高很多。作者提出改进的BH-PREPT算法即LH-PREPT算法。该算法有着兼顾避免级联抢占和抢占优先级代价最小的特点，在通信网络性能方面主要体现在VPN用户的主要业务的时延有所减少。作者根据当前的VPN业务对通信网络性能的需求对总代价计算公式H(l)进行了简单的修正，将y (l)修改成y2(l)以增加优先级抢占代价在总代价中的比重。为了兼顾资源的有效利用和时延特性的提高，，算法在带宽资源和LSP资源条件允许的情况下，用户的通信分组将直接抢占当前的LSP资源以减少时延；当通信资源有限或LSP无法满足用户的通信需求时，算法会将当前的LSP隧道捆绑使用以增加LSP资源的利用率。最后作者使用OPNET仿真了两种算法的通信场景，在两个场景中的通信节点分别采用LH-PREPT算法策略和BH-PREPT算法策略，然后作者对实验结果进行了对比和分析，最后得出了新算法在减少通信时延方面有较好的性能结论。
[Abstract]:The core of multiprotocol label switching is the introduction and application of label concept. Multi-protocol label switching consists of three parts of CE,PE,P, CE is also called user edge routing, mainly for users to access VPN services, PE is also called service provider edge routing, it is directly connected with CE and has the function of LER in MPLS network. At the same time, a virtual route forwarding table is created for each user to isolate multiple VPN users and distribute VPN service labels. P is also called core routing of backbone network. It has the function of LSR in MPLS network and completes fast forwarding and control tasks. DS-TE technology is a collection of differentiated service model and traffic engineering model. In the differentiated service model, the data grouping is divided into different priority according to the service level, which is embodied in the encapsulation of the TOS field, which is composed of six binary digits. It is divided into 12 levels, the highest level belongs to the network control class; The lowest level belongs to the service class as much as possible. When MPLSVPN receives packet with differentiated service level, PE (LER) maps PHB to EXP field in label, converts 12 priority into 8 priority, and then selects different LSP resource according to different priority. The function of traffic engineering model in MPLS VPN is to prevent too many forwarding equivalents of the same type or priority from occupying the same or several LSP resources and causing other bandwidth resources to be wasted. The functional components that MPLS TE needs to use consist of four parts: packet forwarding, information distribution, path calculation and signaling. Therefore, traffic engineering can greatly reduce congestion in VPN networks. Therefore, DS-TE technology is a kind of dual safeguard technology which can not only meet the communication needs of users but also avoid the waste of resources for operators. The starting point of BH algorithm is to solve the problem of bandwidth waste, so the preemption principle and cost formula of the algorithm are mainly based on the utilization of bandwidth. Due to the development of optical communication, the current computer communication network has a very rich bandwidth resources, and the major domestic operators have started to reduce the price of bandwidth services. Therefore, the available bandwidth resources of enterprise users are no longer the bottleneck limiting the quality of communication service. BH-PREPT algorithm has better performance in avoiding cascading preemption. The drawback is that too much attention is paid to the cost of bandwidth waste caused by preemption of LSP, which makes the complexity of the algorithm and the requirement of the device much higher. The author proposes an improved BH-PREPT algorithm, that is, LH-PREPT algorithm. The algorithm has the advantages of avoiding cascade preemption and preemptive priority cost minimization. The performance of communication network is mainly reflected in the reduction of delay of the main services of VPN users. According to the current demand of VPN services for communication network performance, the author modifies the total cost calculation formula (H (l) and modifies the y (l) to y2 (l) to increase the proportion of priority preemptive cost in the total cost. In order to take into account the efficient utilization of resources and the improvement of delay characteristics, when bandwidth resources and LSP resource conditions are allowed, the user's communication packets will directly preempt the current LSP resources to reduce the delay. When the communication resource is limited or the LSP can not meet the user's communication demand, the algorithm will bundle the current LSP tunnel to increase the utilization of the LSP resource. Finally, the author simulates the communication scene of the two algorithms by using OPNET. The communication nodes in the two scenarios adopt the LH-PREPT algorithm strategy and the BH-PREPT algorithm strategy respectively. Then the author compares and analyzes the experimental results. Finally, it is concluded that the new algorithm has good performance in reducing communication delay.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.1

【参考文献】