Facilitating the Implementation of Adaptive Cloud Offloading to Improve the Energy Efficiency of Mobile Applications

Abstract

Cloud offloading -- leveraging remote cloud-based computing resources to execute energy-intensive functionality -- has become a common optimization technique for mobile applications. However, implementing cloud offloading techniques remains a delicate and complex task, reserved for expert programmers. If cloud computing is to realize its promise as a generally applicable, powerful optimization technique for mobile applications, its implementation barrier must be lowered. As we have discovered, reusable system building blocks exposed via a convenient programming model can facilitate the implementation of complex cloud offloading optimizations. This paper describes a system architecture for implementing adaptive cloud offloading optimizations. In particular, the architecture features parameterizable building blocks for monitoring and estimating energy consumption and performance efficiency as well as state synchronization across address spaces, which the mobile programmer can use a la carte. These blocks streamline the implementation procedure for a wide array of adaptive offloading optimizations. Applying this system architecture to third-party mobile applications has optimized their energy efficiency, depending on the execution environment in place.