Cross-layer Energy-Efficient Mobile System Design

Advances in computing technology have resulted in high-end mobile devices with extensive computational capabilities that enable complex applications such as video encoding or lossless data compression. However, the energy requirements of such applications far exceed current battery technology, which is not improving as quickly as computing capabilities. Techniques have been designed targeting each component of the mobile system. However, conserving energy in any single system component in isolation may ultimately result in a net increase in the energy consumption of the whole system because it does not take into account resource usage in different components. To deal with the mismatch between the rapidly increasing energy demands of mobile multimedia systems and the slow increase in battery capacity, research is being focused on energy efficient mobile system design. Through the use of a system energy model that captures the tradeoffs between adaptations in various system components, we design a cross-layer adaptive system, including a MAC protocol, a transport protocol, and a suite of applications. The adaptive MAC layer algorithm, Pincher, achieves energy efficient communication through fast data transmission. Using the mechanisms incorporated in Pincher, we analyze the impact of such adaptations on total system energy consumption. This analysis shows that information about bandwidth must be used by higher layer adaptation algorithms to provide energy efficient adaptations. The adaptive reliability transport protocol, Reaper, supports adaptive applications with various timing and reliability requirements. This transport protocol exposes information about energy costs, data rate available, and expected packet loss rate for use in application adaptations. Finally, the components of the system architecture are tested using adaptive applications. We have developed an adaptive file transfer application and an adaptive video encoding application. This suite of applications is used to show that greater energy savings can be achieved by using a cross-layer adaptive system than by using discrete adaptations in single layers of the system