Ultra-Low-Power Data-Driven Networking System

Power dissipation is a crucial issue for integrating more than billions transistors on a chip. At the same time, reduction in energy consumption will be an essential subject to build future networking environment evolutionarily. This research project proposed in this paper is aiming at developing a data-driven networking system by which the coming networking environment can be realized with ultra-low-power consumption 1/100 to 1/1000 of the present. In order to minimize energy consumption in the system, our project will integrate two approaches by ultimately utilizing passive data-driven principle in realizing both (1)data-driven chip multiprocessor platform based on selftimed elastic pipeline and (2)networking scheme based on ad hoc network technology. As for the platform, particular design considerations by ultimately utilizing passive operation mode in the data-driven principle were taken to achieve both (1)scalable stream execution based on low-power selftimed (clockless) elastic pipelines and (2)realtime multiprocessing without any runtime overheads essentially needed in multimedia communication. As for the networking system, this project introduces energy aware ad hoc networking schemes to reduce the number of control packets that results in reduction of energy consumption at the networking layer. This paper reports the research directions and prospects of our ultra-low-power data-driven networking system project.