With the increasing global stock of vehicles, traffic congestion is becoming more severe and costly in many urban road networks. Road network modeling and optimization are essential tools in predicting traffic flow and reducing network congestion. Markov chains are remarkably capable in modeling complex, dynamic, and large-scale networks; Google's PageRank algorithm is living proof. In this article, we leverage Markov chain theory and its powerful statistical analysis tools to model urban road networks and infer road network performance and traffic congestion patterns, and propose an optimization approach that is based on a Genetic Algorithm to model network-wide optimization decisions. Such decisions target relief from traffic congestion arising from sudden network changes (e.g. rapid increase in vehicles flow, or lane and road closures). The proposed network optimization approach can be used in time-sensitive decision making situations such as crisis response management, where decision time requirements for finding optimal network design to handle such abrupt changes typically don't allow for the traditional agent-based simulation and iterative network design approaches. We detail the mathematical modeling and algorithmic optimization approach and present preliminary results from a sample application.