Software systems for power and energy conservation

The continued scaling of transistors in accordance with Moore's law and the failure of Dennard scaling has resulted in power becoming a critical factor in all facets of microprocessor design. To counter the Power Wall, techniques such as voltage and frequency scaling, heterogeneous hardware and voltage underscaling coupled with improved application reliability have been proposed. Most of the innovation for power conservation has happened at the hardware level. Software needs to be equally responsible for power conservation. For instance, mobile cloud computing is used to offload compute intensive tasks that affect a mobile device's battery. Mobile ad-hoc computing can be used as an alternative to mobile cloud computing in cases where cloud access is not available or is inhibitive to application performance. This thesis presents two systems - Synergy and Equilibria which aid the hardware in its pursuit of conserving power. Synergy is a middleware that increases the battery life for a system of mobile devices connected in a peer-to-peer ad-hoc network. Synergy conserves energy by scaling core frequencies and by intelligently distributing the computation among peer devices. The middleware is not restricted to mobile phones and in no way restricts the mobility of the devices. Equilibria provides software support for voltage domains connected in series. Hardware innovations for power conservation have ignored the power losses incurred by the power delivery circuits in face of low voltage and high current demands. Connecting the voltage domains in a series circuit (instead of parallel) can result in highly efficient power delivery. For series connected voltage domains to work, the voltage draw by each load should be the same. This voltage drop (across the load) is dependent on its CPU utilization. Equilibria is a load balancer which actively monitors the CPU utilization of the processors in the system and ensures equal voltage draw. It runs on a master processor and alters the frequency of the client processors based on the average CPU utilization of all the processors in the system. Equilibria is generic, i.e. the system works irrespective of the software running on the processors