Evaluating computation offloading tradeoffs in mobile cloud computing: a sample application

Abstract

[EN] Mobile cloud computing is generally referred to as the infrastructure where both the data storage and the data processing happen outside of the mobile device. The nature of the connection between the cloud servers and the mobile host are anyway much less reliable than in classical cloud computing with static devices. A compromise must be found between local versus remote computation so to cope with the reduced performance of the data connection and with the characteristics of the mobile device, basically its power availability limitations. In this paper, we evaluate the tradeoffs of computation offloading using as a case study a facial recognition application for smartphones where recognition is a service in the cloud. We present a specifically designed application for mobile devices developed as a component of the proposed evaluation system. The intensive calculus needed for the image manipulation is compared in terms of speed and accuracy both when we delegate it to cloud computing and when we perform it locally on the mobile device. These two alternatives and the intermediate options are compared to determine the optimal settings to take better advantage of integrating these two technologies.