【作者】 马娅婕；

【导师】 杨宗凯；

【作者基本信息】 华中科技大学 ， 通信与信息系统， 2005， 博士

【摘要】 MPLS 技术是将分层网络的第二层交换和第三层路由结合起来的一种L2/L3 集成数据传输技术[1]。在MPLS 网络入口,标签边界路由器处将IP 分组映射为特定的转发等价类FEC,然后再将FEC 用定长的标签编码表示。标签将插入到IP 分组的首部,沿着标签交换路径LSP 的后继节点以分组标签作为索引,查找表示下一跳的新标签,然后用新标签取代旧标签,将分组转发出去,直至标签交换路径的出口。由于MPLS 网络的标签交换特性,使得MPLS 在提高了网络的交换速度的同时,也面临着扩展性方面的问题,具体表现在: 1. MPLS 网络采用标签交换路径LSP 对分组进行转发,LSP 是建立在物理路径之上的逻辑连接,因此若要对所有n 个目的地建立点到点的连接以实现可达性时,网络需要建立O(n2)条LSP,这不利于MPLS 的扩展; 2. 转发等价类对应于不同服务等级和服务质量的要求,对于不同的流量分组,即使它们的源和目的地地址都相同,也可能会使用不同的FEC 来进行分类,相应的,也需要有相同数量的标签来标识。由于标签的长度只有20 比特,因此当网络规模扩大时,MPLS 技术将面临严重的标签匮乏的问题; 3. 由于组播数据流不能象单播数据流一样基于分组目的地进行聚合,而且由于组播数据流的粒度比单播要细得多,因此实现MPLS 的组播技术将面临标签缺乏的问题,这也不利于可扩展性组播协议的研究。为了解决以上三点带来的MPLS 网络的扩展性问题,本论文改变了现有策略中以流量为聚合对象的方法,将MPLS 网络拓扑作为研究对象,对MPLS 网络的拓扑聚合算法进行研究,通过减少出口节点的数目以减少MPLS 网络建立LSP 的数量; 同时减少MPLS 网络内部参与交换的节点的数量以进一步减少标签的消耗。通过这种聚合策略,使MPLS 网络拓扑结构得以简化,提高了网络的可扩展性; 在此聚合后的拓扑之上,设计了MPLS 网络的整体聚合方案,并具体针对组播技术,提出了聚合拓扑层次组播协议,该协议可以使MPLS 减少网络建立组播树的数量,进而减少标签消耗,不更多还原

【Abstract】 MPLS is an integrated transmission technology that combines the switching in layer 2 and routing in layer 3. At the ingress site of MPLS network, the Label Edge Router (LER) maps IP packets into different Forwarding Equivalence Classes (FECs). Then the FEC to which the packet is assigned is encoded as a short fixed length value known as a “label”. The label acts as a shim header that is inserted between the network layer header and the data link layer header. The label will be the only index of the packet which is switched along the Label Switching Path (LSP) and forwarded to its next hop. Then a new label replaces the old one and the packet will be sent out till the egress router of the LSP. The characteristics of the label switching of MPLS speed up the switching rate, as well as it faces the problem of scalability. Several difficulties are itemized as follows. 1. MPLS network forwards packets using LSPs. LSP is a logical point-to-point connection which is set up based on the physical links. So, it needs O(n2) LSPs to make point-to-point connections for all the n destinations in the MPLS network. As the network grows, how to realize the LSP aggregation function is an obstacle when implementing MPLS label switching in a scalable manner. 2. FECs match to different class of services and quality of services. For different traffic flows, even they have the same source and destination addresses, they may need different FECs to classify; accordingly, they need the same quantities of labels as the FECS to identify. As the label length is only 20 bits, when the scale of network enlarges, MPLS will face to a severe problem of lack of labels. 3. Unlike the unicast, whose address aggregation can be coupled with hierarchical address allocation, a multicast address corresponds to a logical group and does not convey any information on the location of its members which means the multicast traffics are very difficult to aggregate. On the other hand, the granularities for the multicast stream are much finer than the unicast and accordingly will consume more labels. So, how to realize the aggregation/merging function is an obstacle when implementing MPLS multicast in a scalable manner. To solve the scalability problems addressed above, focusing on the MPLS network topology, a topology aggregation scheme of the physical MPLS network is investigated. The aggregation scheme reduces the amount of the egress nodes when establishing LSPs. As the result, the amount of the LSPs in MPLS networks is reduced accordingly. Further more, the amount of LSRs that can participate in switching is reduced also to decrease the consumption of the labels. After such aggregation scheme, MPLS topology is simplified. The MPLS multicast protocol is introduced based on the topology aggregation scheme. So it can reduce the amount of multicast trees that the MPLS network sets up, accordingly, the labels will be saved. This scheme can not only satisfy the scalability of MPLS network, but also has higher adaptability to different applications. The works in this thesis has been supported by the National Science Foundation of China “Investigation of Wireless Multimedia Techniques based on Multimedia Transmission Property”(No.60202005) Australian Research Council Foundation “Interactive video-on-demand for e-Learning”(No.LX0240468), and also supported by the project of “Design and Development of Hi-speed Broadband Router-AVIA”that cooperated with COMBRIO Inc. The research of this thesis includes: 1. The characteristics and topology aggregation algorithm of MPLS network edge nodes. A method of constructing minimum weighted dominating set based on the constraint of bandwidth or delay is proposed. This method can be used for aggregating the MPLS edge sub-network topology. A parallel approximation algorithm with O(n) computation complexity and O(Δn) message complexity was designed. This algorithm can generate a weighted dominating set based on single constraint object. It considers the bandwidth between the dominator and dominates. Using the weight, the dominating set has the optimal weighted feature and can satisfy the demands for certain constraints. 2. The aggregation algorithm and its performance of MPLS core network. AFixed-Edge Connected Dominating Set (FECDS) algorithm is designed to aggregate the MPLS core sub-network topology. This algorithm considers the aggregated topology of MPLS edge sub-network. Running the algorithm can generate a connected dominating set of the core sub-network without ignoring the characteristics of the edge sub-network. The resulted whole aggregated topology of MPLS network is a connected entity while keeping the aggregating features of the edge nodes and core node respectively. 3. MPLS overall topology aggregation algorithm. Based on the topology algorithms above, the whole MPLS aggregation strategy is discussed. And the research is focus on the multicast algorithm in aggregated MPLS topology. In order to improve the scalability and reduce the label consumption of MPLS multicast, an Aggregated Topology Hierarchical Multicast (ATHM) protocol is proposed. It constructs the multicast trees based on the aggregation topology achieved in former chapters. The ATHM protocol divides the MPLS area into different sub-areas to form 2-layer hierarchical multicast area according to the dominators and dominatees in the generated aggregated topology. At the same time, the label stack is used to save the label space. 4. The implementation of aggregatable MPLS model in high-speed broadband router based on embedded Linux. Cooperating with the project “Design and Development of Hi-speed Broadband Router-AVIA”, the MPLS topology aggregation scheme modules are designed and implemented in Linux platform.更多还原