Expanding beyond the traditional focus on performance, power management is now a well established area in computer systems research. This is specially true in the embedded arena, but also in servers; it comes from the desire to extend device lifetime (e.g., flying across the Pacific in a single battery) and/or cost of energy (e.g., paying less than 15% of total costs for energy in data centers). In this arena, the trend is a paradigm shift from a maximum performance approach to a power/performance tradeoff.
   However, reducing power and energy does not come for free and is not suitable for all systems. Some systems are amenable to such tradeoffs, specially systems where desired properties are well specified. Real-time systems research provides a framework for specifying such requirements, specially the timing requirements of an application. I contend that every application domain is real-time, or else there is virtually no need to execute the application. Some domains are hard real-time (such as mission critical applications, where deadline misses result in system failure), while other domains are soft real-time, with statistically specified performance requirements. As low-power and high-performance are typically conflicting goals, I argue that real-time constraints and power management issues should be analyzed in a unified framework.
   A third issue is that of system utility (also called reward, quality of service QoS or value) and adaptivity. Ideally, systems monitor themselves and adapt the resource allocation to applications, aware of application utilities. Analyzing the issues of power management, real-time constraints and utility within a unified framework has been the subject of my Ph.D. research.