Many of the systems we rely on in our daily lives, such as telecommunications, transportation, and electric power, are designed and built as networks. Our dependence on these infrastructure systems is so great that a large-scale breakdown of any one of them would be catastrophic. With the growth of the Internet and the advent of a common "information infrastructure", there has been an ongoing drive to increase the connectivity within and between these systems. Although this trend has yielded many new efficiencies for daily life, it has come at a cost of greater overall system complexity. In the last decade, most of these infrastructure systems have experienced major disruptions of some form, and many researchers and policy makers are questioning the extent to which the perceived fragility for these systems is a direct consequence of their interconnected nature.
In this talk, I will provide a framework for thinking about cascading failures. Leveraging notions from network optimization and distributed control, I will show how simple models of unicommodity flows can be effective for understanding many of the key tradeoffs and tensions in the operation, control, and design of network systems. I will outline some of the outstanding research questions for this area, and describe an ongoing case study with the Union Pacific Railroad.
Copyright. All Rights Reserved.