Energy Efficient System Design and Utilization a Dissertation Submitted to the Department of Electrical Engineering and the Committee on Graduate Studies of Stanford University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Energy consumption of electronic devices has become a serious concern in recent years. Energy efficiency is necessary to lengthen the battery lifetime in portable systems, as well as to reduce the operational costs (e.g. cost of electricity) and the environmental impact (e.g. cooling fan noise) of stationary systems. Optimization in design and utilization of both hardware and software is needed in order to achieve more energy efficient systems. In this thesis I first discuss power management algorithms that enable optimal utilization of hardware at run time. Next, I discuss the new dynamic voltage scaling algorithm that complements power management by scaling processor frequency and voltage depending on the needs of the system. Finally, I develop a modular approach for design and simulation of hardware and software energy consumption at the system level. Dynamic power management (DPM) algorithms aim to reduce the energy consumption at the system level by selectively placing components into low-power states. Two power management algorithms will be presented that have been derived from the stochastic models of several realistic examples. Both algorithms are event-driven and give optimal results verified by measurements. The first approach is based on renewal theory. This model assumes that the decision to transition to low power state can be made in only one state. Another method developed is based on the Time-Indexed Semi-Markov Decision Process model (TISMDP). TISMDP model assumes that a decision to transition into a lower-power state can be made upon each event occurrence from any number of states. On the other hand, it is also more complex than the approach based on renewal theory. It is important to note that the results obtained by renewal model are guaranteed to match results obtained by TISMDP model, as both approaches give globally optimal solutions. I implemented both power management algorithms on two different classes of devices: two different hard