Reservation-based resource scheduling and code partition in mobile cloud computing

Abstract

Mobile cloudlet has shown great potential to extend the computational limitation of mobile devices. The overall performance of the mobile applications could be improved when the computational gain from the cloudlet overweights the transmission cost. Yet, resources scheduling on the cloudlet is a challenging problem when there are multiple mobile offloading requests at the same time. Since the offloading requests would compete with each other and each offloading decision would impact the others, finding the optimal execution layout is quite difficult. In this paper, we propose a joint resource scheduling and code partition scheme that can efficiently allocate cloudlet resources to multiple mobile users. Our goal is to maximize the cloudlet throughput as well as to reduce the mobile application's execution time. Unlike the current approach which considers an infinite resource limitation on the cloud, our algorithm dynamically allocate the cloudlet resource based on the current usage. Both real world experiments and trace-driven simulations show that our solution can improve the cloudlet throughput by 20%–35% in the meantime provide speedup to mobile devices.