Topology Measurement And Fault Detection For Computer Networks:Methodlogy Study And System Development

Network management becomes more and more important for today’s Internet, and the Internet’s topology plays a key role in the Internet’s fault avoidance, configuration and performance management, Internet upgrade and optimization. Thus complete and accurate network topology is the basis for network management. Concurrent with the increasing diversity, transparency, dynamics of network devices, so discovering network topology becomes increasing challenging. In order to obtain target network router level topology, we propose a novel multi-source topology mapping methodology based TraceNET in this dissertation. Existing works on link-layer topology discovery is based on the assumption that the network is static without mobile devices. However with popularity of wireless access, there are more and more mobile devices, making the network more and more dynamic. Thus an algorithm for discovering the link-layer topology under dynamic networking environments was proposed in this dissertation. In order to keep a smooth real-time network, should efficiently and accurately to locate, detect and exclude network faults. So in this dissertation, we design and implement a system to detect Ethernet normal faults based on SNMP and MIB-2.The main contributions of this dissertation are listed as following:First, a novel multi-source topology mapping methodology based TraceNET for discovering router level topology is proposed, and apply this approach for mapping the topologies of the campus network of USTC and CERNET. The measurement results suggest that this method solves the router IP alias resolution and the subnet discovery problems very efficiently, and accurately obtains the network topologies in terms of routers and subnets, therefore providing valuable topological information on the Internet for network management and study.Second, an algorithm for discovering link-layer topology under dynamic networking environments is proposed. Simulation experiments on ns-3and the real-world networking topology measurement results indicate that this algorithm could exclude the impact of mobile devices and discover the linker-layer topologies efficiently and accurately.Last, a system for detecting Ethernet normal faults is designed and implemented based SNMP and MIB-2. Simulation experimental results on the ns-3platform and the real-world networking faults monitoring results indicate that this method which used in the system could discovery and locate Ethernet faults with high efficiency and accuracy.