SHF: Small: Application Shrinking for Reducing Energy Consumption

University Of California-San Diego, La Jolla CA

Investigators

Abstract

The management of energy consumption is a key driver in the design of modern computing platforms, from mobile devices (where lower consumption leads to better battery lifetimes) to desktop and servers (where lower consumption leads to significant decreases in electricity costs and heat dissipation). However, despite numerous advances in low power design technologies, the power consumption of the IT industry remains significant and worse, is slated to grow in the coming years. The prospect of growing energy demands for computing systems has implications for the economy, the environment and for society in general. The broad goal of this research is to make it easier to reduce the energy consumption of computer systems, while still maintaining their functionality. An emerging way to reduce energy consumption is to leverage heterogeneity in the choice of individual components to tradeoff power consumption versus performance. The key idea of such "Collaborative Heterogeneous" systems is to use low-power low-performance helper platforms in lieu of high-performance high-power parts whenever feasible, thus keeping the high-power subsystems shutdown to save energy. To unleash the full potential of Collaborate Heterogeneous systems, this research will develop techniques that can automatically generate shrunk versions of applications that are suitable for running on low-power resource-constrained helper platforms. This research will lead to new techniques for application shrinkage, to better tools for measuring energy savings, and to an empirical evaluation of the energy savings made possible by Collaborative Heterogeneous systems. The techniques developed in this research will also be useful for generating "shrunk" versions of applications for use in heterogeneous multi-core Architectures and for running on portable energy-constrained devices like SmartPhones.