JouleGuard: Energy Guarantees for Approximate Applications

Energy consumption limits battery life in mobile devices and increases costs for servers and data centers. Approximate computing addresses energy concerns by allowing applica-tions to trade accuracy for decreased energy consumption. Approximation frameworks can guarantee accuracy or per-formance and generally minimize energy; however, they pro-vide no energy guarantees. Such guarantees would be bene-ficial for users who have a fixed energy budget and want to maximize application accuracy within that budget. We ad-dress this need by presenting JouleGuard: a runtime con-trol system that coordinates approximate applications with system resource usage to provide control theoretic formal guarantees of energy consumption, while maximizing accu-racy. We implement JouleGuard and test it on three differ-ent platforms (a mobile, tablet, and server) with eight dif-ferent approximate applications created from two different frameworks. We find that JouleGuard respects energy bud-gets, provides near optimal accuracy, adapts to phases in ap-plication workload, and provides better outcomes than appli-cation approximation or system resource adaptation alone. JouleGuard is general with respect to the applications and systems it controls, making it a suitable runtime for a num-ber of approximate computing frameworks