Bridging the Energy-Efficiency Gap in a Future of Massive Data

Fred Chong
Director, Greenscale Center for Energy-Efficient Computing
Director, Computer Engineering
University of California at Santa Barbara

In an era when massive data will enable unprecedented opportunities in business and science, computing faces significant challenges in the scaling of performance and energy consumption. The overarching problem is that of exponentially growing, massive global data, with 1,000X growth within the next 13 years. The World Economic Forum recently focused on the potential opportunities from exploiting this increasing flood of global data. Unfortunately, the International Technology Roadmap for Semiconductors (ITRS) predicts transistor density doubling only every three years. Under these assumptions, technology scaling will provide no more than 25X improved computational efficiency in the same 13 years, leaving at least a 40X gap between computation growth and data growth. The actual gap is likely to be far worse, as transistor energy efficiency has begun to improve more slowly than density. Without significant improvements in energy efficiency, tomorrow's computing infrastructure will either require substantially more investment and energy or fail to realize the value of the world's data.

In this talk, I will describe several research directions that strive to sustain the energy-efficiency improvements necessary to keep pace with data growth. These range in focus from device technology to system design, and from mobile to warehouse-scale systems. I will also discuss metrics that better capture emerging issues of green versus brown energy, managing manufacturing impact and IT lifecycles, and water and other resource usage.