Our nation's security as well as the quality of life of its citizenry depends on the continuous reliable operation of a collection of complicated interdependent infrastructures including transportation, electric power, oil, gas, telecommunications and emergency services. A disruption in one infrastructure can quickly and significantly impact another, causing ripples across the nation. Our infrastructures are increasingly reliant on new information technologies and the Internet to operate, often being connected to one another via electronic, informational links. While these technologies allow for enormous gains in efficiency, they also create new vulnerabilities. The same technology that allows us to transmit information around the globe at the click of a mouse can be used to disrupt our vital systems including the flow of electric power or water, and the dispatch of emergency services. Many of our infrastructures are privately-owned by companies that face strong pressure for short-term financial results. This pressure may lead to under-investment for the future. The confluence of increasing interdependency among infrastructures and increasing economic pressures implies a significant need to understand how these systems are interdependent so that systems of infrastructures, or "mega infrastructures," are less vulnerable to disruption while maintaining their financial viability. This project focuses on the development of computationally tractable mathematical models to asse the performance of interdependent infrastructure and to optimize investment. A key aspect of the mathematical approach we are taking is the creation of a unifying mathematical framework to represent these mega infrastructures and a collection of algorithms that can be used to estimate performance and optimize investment. This unifying mathematical framework needs to be of sufficient detail to provide an understanding of the impact of the physical facilities and cyber technologies on performance, but remain computationally tractable for realistic application. It must also play explicit attention to the underlying uncertainty in these systems. A primary focus of this effort is the applicability of Markov and semi-Markov models for this purpose.