Mobile cloud face recognition based on smart cloud ranking

Abstract

A new trend has been explored to tackle the limited resources of mobile devices: mobile cloud computing. However, there is a difficulty in deciding--at runtime--the more appropriate target resources where to run offloaded mobile device tasks. This study presents SmartRank, a scheduling approach to perform load partitioning and offloading for mobile applications using cloud computing to increase performance in terms of response time. We have applied the approach to a face recognition process based on cloudlet federation and resource ranking through balanced metrics. Besides, the tool was evaluated by two ways. First, by using system modeling (continuous-time Markov chain). Second, by using a full factorial experimental design to calibrate the SmartRank with the most suitable partitioning decision. Nevertheless, SmartRank uses an equation that is extensible to include new parameters and make it applicable to other scenarios.