Characterize energy impact of concurrent network-intensive applications on mobile platforms

The cellular network bandwidth increases significantly in the past few years, stimulated by many popular network-intensive applications, such as video streaming and cloud storage usages. Meanwhile, more and more users enjoy the multitasking feature of mobile devices and concurrently run a number of applications. Given these two trends and the fact that extended battery life remains to be a critical factor for small form factor devices, e.g. smartphones and tablets, it is imperative to understand the energy impact of multiple applications running concurrently on such platforms.

In this paper, we characterize and understand the energy and performance impact of concurrent applications via a comprehensive set of carefully designed experiments. Specifically, we focus on network-intensive applications since most usage models today are driven by always-on communication activities. We make several significant contributions to shed light on understanding the energy behavior of concurrent applications. Firstly, we find out that running multiple network-intensive applications concurrently can significantly improve energy efficiency, up to 2.2X compared to running them separately. Secondly, we observe that power consumption from CPU and System on Chip (SoC) are the primary culprits of power dynamic for network-intensive applications; while communication components, including Network Interface Card (NIC), poses very little power consumption variation with different throughput. Thirdly, we investigate, in detail, the significant impact of signal strength on the energy consumption and throughput performance. Our findings and analysis can be applied to provide helpful guidance for a wide range of research aiming to optimize mobile device energy efficiency, e.g. transmission scheduling and protocol design in cellular networks.