Label Switching And Time Synchronization Technologies In Packet Transport Networks

With the explosive increase of high speed Internet, video services and multimedia applications, IP-based data traffic is occupying more and more bandwidth in the modern telecommunication network. It becomes quite clear that the structure of the traditional telecommunication network, which is designed for carrying mainly voice services, needs to be reengineered to handle data traffic in a more efficient way. The structure of the telecommunication network is changing from a TDM-based network to Carrier Ethernet and PTN (Packet Transport Network). The new structure inherits the advantages in the control, management and survivability aspects of the traditional telecommunication network, and is more suitable for carrying IP-based data traffic. In the new structure, multiple services are transported in the common network. The detailed contents of this dissertation include of realizing the agile allocation of the capacity and resources, increasing the intelligence of the packet transport network in the aspects of control and management planes, and decreasing the transport cost per bit in the new structure.Based on the analyses of the requirements of the telecommunication networks, this dissertation demonstrates technologies about the control plane and the label distribution of the MPLS-TP (Multi-Protocol Label Switching Transport Profile) network, and the time synchronization in PTN. The contributions of this dissertation mainly can be summarized as follows.(1) In the part of the control plane of MPLS-TP, an improved delay and bandwidth constrained path computation algorithm with traffic engineering objectives is proposed for the requirement of effective QoS routing mechanisms, and a risk-sharing multicast recovery scheme based on redundant trees is proposed for the requirement of resilient multicast mechanisms. The former takes both the bandwidth utilization ratio and the resident bandwidth on the links into consideration in the computation of the route, and can balance the load of the network. The later proposes an improved primary tree selection method to enhance the redundant tree recovery method in multipoint optical networks. This method not only can ensure the minimum hop distance from the source to every destination, but also can make nodes in the network share the risk of failure and degradation by averaging the number of messages triggered.(2) In the part of the label distribution of MPLS-TP, a sign-label based mechanism is proposed firstly to enable the communication between a Maintenance End Point (MEP) and a Maintenance Intermediate Point (MIP) in the MPLS-TP OAM (Operations, Administration, and Maintenance) architecture. This mechanism realizes the node numbering in the point-to-point co-routed bidirectional path of MPLS-TP networks, and the binding of the forward and the backward directions of the bidirectional transport path. Thus, it supplies a return path for the OAM communication without the need to support IP-based routing and forwarding. Secondly, a symmetric label distribution mechanism is proposed, which is proved to be available in NS2. Compared with the existing methods, the proposed mechanism is more efficient to construct bidirectional paths in MPLS-TP networks. It can ensure that the nodes along the path know the pairing relationship of the forward and backward directions of the transport path, and thus the nodes can be used as end points to protect all or part of that path.(3) In the part of the time synchronization of PTN, an enhanced end-to-end Transparent Clock (TC) mechanism with a fixed delay ratio and another one based on syntonized PTN are proposed. IEEE1588assumes symmetrical paths, and asymmetry of the message propagation delay may cause large errors in clock synchronization. As this standard is adopted in various applications such as telecommunications backhaul networks, it is extended to enable correction for asymmetry by measuring peer delay and residence time in transparent clock devices. However, these measurements are based on local clocks, and errors resulting from differences in the rates of the master and the transparent clock devices affect the synchronization accuracy. Two methods are proposed to solve the rate mismatch problem, and they both need two rounds of message exchanges. In the first method, the PTP massages in the second round are delayed for a specific period of time related to the fixed delay ratio. In the second method, it is assumed that the transparent clocks have been syntonized. The second method is compatible with current IEEE1588systems, and do not need to change the process of the PTP messages in the transparent clocks. By theoretical analyses and computer simulations, the proposed mechanisms are proved to be able to get an accurate recovered time and support the multi-operator scenario efficiently.