Research On Technologies Of Multi-service Ring Protection And Strategies On Network Optimization In Carrier Ethernet Networks

Carrier Ethernet technologies provide Ethernet with the abilities of the carrier class scalbility, manageability and QoS(Quality of Service) guarantee. The deployment of Ethernet into carrier networks drives the transformation from traditional networks into packet networks. However, the reliability issues are hard to be resolved efficiently as a result of the packet characteristics and mechanisms of Ethernet. And the issues impact the large scale deployment of Ethenet into carrier networks.The traditional rigid protection is not good for providing differential protection strategy nor optimizing the network resource. On the other hand, resilient protection brings the uncertain of QoP (Quality of Protection). A QoP model based on MSR(Multiple Services Ring) uses the available bandwidth before and after fault to measure the QoP. And the model introduces the QoP factor to quantitatively evaluate the QoP. Based on this model, a revernue-oriented MQoP mechanism can provide differential protection for multiple services according to the different requirements of protection and revernue of services. The analysis of the experiment results on the system platform shows that the differential QoP for multiple services can be quantitatively guaranteed in the MSR resilient pipe through the mechanism.The off-the-shelf protection mechanisms can neither exerte the simple and efficient characteristics of Ethernet nor satifify the carrier class50ms protection switching requirement or have the limitations of topology structures. A pre-configure R-cycle (Resilient cycle) technology and a RER(Resilient Ethernet Ring) protection control protocol based on R-cycle can achieve fast protection in any topology without any change of the Ethernet forwarding method. The technology and protocol adopt distributed detection and control mechanism and implement the simple and efficient automatic protection switching operations within the R-cycles. Regarding the share links ownership problem, a constraint on the number of node adjacent link ownership with R-cycle can avoid the super loop in network planning. Through the comparative validation with the ring protection protocols on the commercial systems, it can be proved that RER protocol can achive distributed carrier class50ms fast protection switching and have much more advantages on the scalbility and universality relative to other ring protection protocols. The protection protocols based on the Ethernet layer require all the nodes in the network flush their address tables after failure occurses. The traffic flooding resulted by the operations will impact the service recovery time extremely. The available solution need to introduce a mount of protocol overhead and occupy a lot of bandwidth. A PBTF(Pathd based Trigger Flushing) mechanism can decrease the number of node which need to flush address table efficiently without increasing protocol overhead. By comparing with the newest protocol standard theoretically and experimentally, it can be proved that the mechanism can avoid the large-scale traffic flooding problem as a result of all the nodes flushing operation in the network. The comparative data of the experiments on the commercial systems show that RER protocol with PBTF mechanism is better than G.8032protocol regaring to the service recovery time. And the effect is more and more obvious as the network scale expanding and the traffic increasing. The data of the experiments show that the PBTF mechanism can shorten50%of the recovery time under the situation of more traffic(beyond16k) in the classic three level ring networks of metropolitan area network.Different hierarchical between rings in mesh networks and different block port positions will lead to different bandwidth requirements and traffic allocation. In order to optimize the network structure and traffic allocation, a network planning model based on the R-cycle can analyse and design the objectives of link bandwidth, spare bandwidth allocation, load balance, minimum cost and maximum revernue as well as the corresponding constraints. A heuristic HRNPA network planning algorithm and its sub-algorithms can complete the computation of planning problem in polynomial time. Through comparing with the best results obtained by the enumeration in the simulation, it can be showed that the difference between the optimal results obtained by HRNPA and the best result are less than10%.