Research On Survivability Inter-Domain Routing Protocols In Internet

Internet is increasingly becoming the communication infrastructure and starting to carry more loss and delay-sensitive mission-critical traffic, and those mission-critical services depend on the assumption that network being uninterrupted where even very short-lived disruption can result unacceptable losses in some important national defense, economic cases, which manifest the importance of network survivability. Routing protocols have the most significant impact on data plane performance and network stability for given the network topology and architecture. Whereas, extensive measurements on Internet have showed that routing failures are very widespread, even multi-homing AS does not always provide more fault tolerance than it should. Many prior studies have revealed that the Internet routing infrastructure is highly redundant; the ideal routing protocols should provide continuous end-to-end communications as long as the underlying topology remains connected. Nevertheless, the de facto inter-domain routing protocol BGP, the topology changes can trigger reactive convergence on a new globally consistent route views, the convergence time even lasted thirty minutes at the worst case, and transient failure during the routing convergence degraded the forwarding performance of significantly. Furthermore, BGP may lead to inter-domain churn, which compromises the stability of the network.The analysis on the difficulties faced by BGP shows that both eBGP and iBGP would cause the AS-level and router-level transient failure. Specially, the iBGP may cause lots of unnecessary inter-domain churn. This thesis points out these problems stem from the BGP path propagation mechanism, sensitive failure handling mechanism and ignoring the interaction of eBGP/iBGP. Therefore, the main theme for this thesis has been to develop inter-domain routing protocols that offer fast recovery and churn reduction. After analyzing the fundamental mechanism, characteristics, performance of existing representative fast recovery routing protocols and churn reduction technologies. To address the shortcomings in current researches, this thesis makes an in-depth study on existing problems to be solved in BGP.Considering the lack of AS-level path diversity and transient failure caused by eBGP, the thesis first focuses on the eBGP for addressing the inefficiency of current pre-computed backup-path approaches, including its unable to ensure path diversity, the forwarding loop caused by inconsistent backup path decision, the complex signal and storage overhead.We propose a variant of BGP with robust failure tolerant by using Multiple Forwarding Tree, named MFT2-BGP. MFT2-BGP assigns path identifier for every path when announcing prefix, and constructs policy-compliant multiple forwarding trees with low message overhead. It can definitely guarantee every AS node learn multiple forwarding path, and keep all nodes on the forwarding tree corresponding to specific path identifier with consistent routing view so that it becomes easier to deal with multiple failure and deployment. Moreover, allowing packet to dynamically switch forwarding tree, MFT2-BGP can achieve disruption-free forwarding in the presence of transient failure. Besides that, MFT2-BGP also reduces the convergence time by carrying"root cause notification"in UPDATE message and reduces routing system churn by suppress transient failure. The extensive simulations demonstrate that MFT2-BGP can reduce convergence time, alleviate forwarding disruption caused by link failure, and strengthen the routing system stability.We propose a lightweight backup AS-Address Advertisement with Root Cause Notification-BGP (named BAR-BGP) inter-domain routing protocol for achieving fast recovery. Instead of pre-computing failover paths, it deals with transient failure by combining the method of backup AS advertisement and reactive re-computing feasible paths. When transient failure occurred at an AS node, BAR-BGP will encapsulate and forward packets to backup AS, and use recovery forwarding mode to re-compute new feasible forwarding path in light of local failure and failue-carrying packet for reducing loss of connectivity further. The detailed simulation demonstrates that BAR-BGP effectively reduces forwarding disruption caused by the transient failure without paying cost to maintain backup AS-Address.Considering the current research widely ignoring the iBGP impact on network performance, the thesis concludes that the iBGP is not able to ensure router-level path diversity, route reflection will bring several negative impacts on the overall routing system such as iBGP path exploration, UPDATE duplication, and BGP churn. The thesis proposed two approaches aiming at the lacks of next hop path diversity and too sensitive failure responding mechanism in iBGP to offer fast recovery and churn reduction.We propose Multipath Route Control Platform (named MP-RCP) architecture and protocol based on RCP. MP-RCP decouples the routing selection and dissemination, and it also logically separates the client router's route table, MP-RCP takes advantage of global visible route to assign mutlipath for every router in the AS. MP-RCP also decouples the failure notification and recovery, and the MP-RCP is responsible for failure responsive in light of the local AS policy. The detailed theoretical analysis and simulation demonstrates that MP-RCP effectively reduces forwarding disruption and BGP churn.We propose a variant of iBGP based on egress path identifier (EID), named EID-iBGP. EID-iBGP eliminates iBGP path exploration with the adding EID in UPDATE in redundancy route reflection. Furthermore, we propose Add-EID-iBGP allowing propagate multiple EID without introducing of multipath storage overhead, Add-EID-iBGP explicitly makes use of EID tags to localize the failure information avoiding unnecessary BGP churn, and it reduces forwarding disruption by allowing traffic to switch different virtualization network corresponding to EID. The thesis also discuss the WRATE settings according to the iBGP convergence time analysis. The detailed theoretical analysis and simulation demonstrates that Add-EID-iBGP effectively reduces iBGP path exploration, forwarding disruption and BGP churn.