Optimization algorithms for energy-efficient data centers

Abstract

Real-time optimization algorithms for managing energy efficiency in data centers have been developed and implemented. For example, for a given cooling configuration (which is being measured and modeled in real-time using IBM's Measurement and Management Technologies) an optimization algorithm allows identifying the optimum placement of new servers (or workloads) or alternatively where to remove servers (or workload) for different constraints. Another optimization algorithm optimizes performance of the data center without creating hotspots. The optimization algorithms use a physical model in junction with linear programming as well as linear least square methods. © 2014 ASME.