Network configurations are typically done by manually setting up the configuration files. Such manual configuration requires the network operators to be able to deal with low-level protocol-specific parameters as well as have a big picture of the network in mind. Thus, network configuration is prone to error. What makes things worse is that the network is a dynamic system where there are link failures/additions and routing announcements from outside the network. The growth of the network size and the huge number of routing events make it hard to predict the effect of the configuration. In this thesis, we proposed a graph-algorithm-based approach to verify the network configuration robustness. In addition to verifying properties for one network snapshot, our approach can also verify these properties under intra-domain and inter-domain routing events efficiently. We implement the proposed approach and evaluate under both synthe- sized and real network configurations. The experiment results show that our approach can achieve at least 100x speedup over the existing approaches when verifying properties under routing events and scale to large networks with thousands of nodes.