Characterizing and improving energy-delay tradeoffs in heterogeneous communication systems

Abstract

Communication systems with multiple computing resources are gaining popularity. These devices consume a considerable amount of energy, and require that the system resources be fully utilized at all times. In this paper, we aim to reduce the energy consumption and also improve the system utilization by incorporating performance-enhancement and power-optimization techniques into a multi-resource heterogeneous communication system. We study the performance and energy improvements contributed by our techniques, using a traditional evaluation framework. However, this framework fails to clearly model the performance-energy tradeoffs in the system. Hence, we propose a more relevant framework with a new metric named Energy-resource efficiency (ERE) to study these systems. ERE defines a link between the performance and energy variations in a system to clearly highlight the various performance-energy tradeoffs. Our experimental results shown that, when ERE is used to calibrate the performance-energy tradeoffs, one can achieve up to 80% performance and energy gains in a power-aware heterogeneous communication system.