Research On MPLS And Its Network Optimization Algorithms

With the development of Internet, traffic in network increased with geometric series. Further more, different kinds of traffic ask for discrepant level QoS. Amechanism is necessary for network to assign traffic onto physical link and makeresources utility optimization. But traditional routing algorithms based on SPF don'thave this function. "Traffic Enginerring" is a kind of network control mechanism which can balance the traffic in network and optimize the resources utility.The concept "Traffic Engineering" is derived from the methods of the integrating administration to network resources and traffic, which can optimize the performance of network, improve the availability and the cost-efficiency-ratio, maximize the utilization efficiency of the network resources, and get the best validity and reliability of the network operation. So, TE become the basic demand for the further development of large ISP. IETF proposed Multiprotocol Label Switching (MPLS) to solve this problem.Multiprotocol Label Switching add the connect-oriented mechanism to the unconnected IP protocol, and uses rapid switching in the middle of the network instead of traditional routing. So MPLS is thought as the best technology to implement "Traffic Engineering".This paper emphasises on MPLS TE mechanism and the routing method of it. Forthermore, the routing method is discussed and studied in detail. A method using Genetic Algorithm to solve the problem of network resource optimized configuring in the environment of MPLS has been discussed emphatically. We also build mathematic models, which can be easily programed into modularized software objects, for all the MPLS network components that are involved in MPLS network resource optimized configuring. And it is important to build clear and effective model objects for implementing optimizing algorithm. Using Genetic Algorthm in network optimizing, we get a good optimized effect. Genetic Algorthm can be modularized easily so that the network optimizing algorithm can be built with clear structure. And implicit parallelism property inhereing in Genetic Algorthm accords with high compute efficiency for network optimizing. In this paper, we develop the traditonal Dijkstra algorithm, replacing the weight by hop, delay, bandwidth and cost to route. And take the results as the candidate population to optimize with GA. Finally, results a minimummaximum-bandwidth-efficiency scheme.