Assessing the benefits of computational offloading in mobile-cloud applications.

ABSTRACT This paper presents the results of a formative study conducted to determine the effects of computation offloading in mobile applications by comparing “application performance” (chiefly energy consumption and response time). The study examined two general execution scenarios: (1) computation is performed locally on a mobile device, and (2) when it is offloaded entirely to the cloud. The study also carefully considered the underlying network characteristics as an important factor affecting the performance. More specifically, we refactored two mobile applications to offload their computationally intensive functionality to execute in the cloud. We then profiled these applications under different network conditions, and carefully measured “application performance” in each case. The results were not as conclusive as we had expected. On fast networks, offloading is almost always beneficial. However, on slower networks, the offloading cost-benefit analysis is not as clear cut. The characteristics of the data transferred between the mobile device and the cloud may be a deciding factor in determining whether offloading a computation would improve performance.