| The development of Internet has put a great impact on the human society.The role it played in social activities and daily life is becoming more important than before.However, it can also cause great damage even making the society paralytic if a fatal network failure happened. Network survivability mainly concentrate on fault tolerance and intrusion, which specifically refers to the capability of fulfill its mission in a timely manner,in the presence of attacks,failures or accidents, including failure detection,failure location,failure notification and failure recovery.The survivability, especially in multi-layer networks,has become a key technology for layered network. The overlay network built above the IP Network is the typical one which uses the multi-layer network for improving the network preformance. It has been proposed as an alternative method to overcome the IP Network anomalies.There are two general classes of overlay network,routing overlay and storage and look overlay.Routing overlay has routing function,such as ron;Storage and look overlay uses for distributed system,such as p2p network.(If not specified,this article referred to overlay network is routing overlay.) The paper indicates some aspects of overlay network exactly, including sort of platform, topology structure, security mechanism and route algorithm. Some protocols about network failure detection, such as hello protocol in ospf protocols and bidirectional forwarding detection, are also included.Three aspects of overlay network has been researched in the paper, which are link failure detection model, distinguishing method for node failure and link failure, and dual rerouting recovery coordination mechanism.The main contributions of the paper are list as follows:1. First, we carry out a detailed analysis of overlay network link failures, including link failure and link congestion. Next, we propose an improved link failure detection model based on the existing model. The simulation of the model proves that it can reduce the number of sending packages and make the number of detected link failures closer to the real failure number.2. Secondly,overlay network failure includes node failure and link failure,which can cause different network abnormalities, so they should be distinguished from each other for the efficient recovery. In this paper ,we propose a distinction way of link failure and node failure. The method can effectively distinguish the two types of failure,and informe that the link failure shouled be recovered by reroutin. For node failure, it is best to broadcast in the network so that other nodes can stop sending packets to the failure node for reducing network congestion.3. Finally, we study the problems of dual reroute between overlay layer and IP layer in face of IP link failures, suggesting to coordinate the two layer's reroute recovery mechanism. If the network doesn't adopt any coordinate method and let the two recovery mechanism run independently, the recovery conflict will occur and waste the network resource. Therefore, we propose a recovery coordination mechanism based on different Autonomous Systems.It doesn't have to assume which layer detected network failure first so that we can adopt a unified coordination mechanism to deal with the network link failure. Distinguishing between link failure occurred in intra-domain or inter-domain, can provide a more reasonable recovery method. Because the overlay network itself has the advantages of restoring the inter-domain failure, the link failure occurred in inter-domain is better to be restored by the Overlay network recovery mechanism.As to the intra-domain link failure, IP networks are generally able to find a more reasonable alternative route.The simulation show that it has smaller cost rerouting to adopt the dual rerouting coordination mechanism we proposed than IP rerouting mechanism。... |
PDF Full Text Request