It is widely appreciated that small changes in chilled water plant configuration can have large impacts on overall system efficiency, yet finding the optimal control set points for a given condition is nontrivial. Achieving global optimality requires simultaneous consideration all aspects of system operation. For a given load and ambient condition there exists a plant configuration that maximizes the coefficient of performance (COP); this configuration changes over time, necessitating continuous adjustment. The purpose of our project is to demonstrate such a holistic, model-based approach to modeling, optimization, and control at a Verizon data center.