Research On The Fast Failure Recovery Technologies Of IP Networks

Today, network has been integrated into all aspects of people’s daily life. Withthe development of Internet, lots of real time applications such as VoIP, online games,electronic commerce, web video emerge and prosper. In this situation, there is ahigher demand of the IP routing to recovery from failure. Although traditionalrouting protocols can solve almost all types of network failure problems, they willtake up too much time, they may also result in loss and delay of packets, routingloops, link congestions and so on before the completion of re-convergence. Thesecan not meet the demands of real time applications in the network. This thesis willresearch on fast recovery from failures of IP networks.The two existing failure recovery technologies are Reactive and Pro-active, thelatter can significantly shorten the delay time and reduce the packet loss rate, whichcompared with the former. So it is very suitable to deal with the frequent transientfailures and has more practical value. This thesis mainly researches on algorithms offailure fast recovery based on pro-active technology and includes the followingaspects:1. Reasearch the pro-active fast recovery technology and propose IBP algorithm,which is an improved method aims at the network fast recoveries from single failure.The algorithm compares the original network state with the state after failure, thefailure component is always splits the shortest path tree into multiple non-connectedtrees, according to this the algorithm will generates a backup path which couldre-connected those partial trees and then gives the loop-free proof of the backup path.Simulation results show that the average length of the backup path which isgenerated by IBP algorithm is superior to existing algorithms, and the difference issmall compared with the shortest path length, the number of packets loss is alsosignificantly reduced during the network recoveries from the failure, what’s more,the convergence of the IBP algorithm can be guaranteed, too. So it is more suitablefor large-scale network environment.2. The thesis analyses the problem of link congestion results from flowdistribution imbalance and also presents a load balancing method. That is setting thenetwork traffic matrix and objective function to compute an approximating optimum link weights by genetic algorithm. Through optimizing weights to distribute networkflow, and combined with Pro-active failure recovery algorithm, then it can be used torecovery from single failure and also avoid congestions. So this method enhances thefeasibility of failure recovery algorithms.3. Reasearch the algorithms of RRL and MRC, which are used in the situationof fast recovery from multiple failures of network, and propose an improvedalgorithm based on the minimum spanning tree. The method generates the minimumspanning tree of the original network topology by setting an appropriate group oflink weights, resulting in a backup topology. Then adjusting weights of some links toisolate the remaining links and nodes as much as possible, such a cycle, generatingfewer backup topologies, but can also deal with the situation that multiple failureshappen at the same time. Simulation results show that the improved algorithm savesthe storage resources and is suitable for large-scale network environment.