An Investigation of Different Computing Sources for Mobile Application Outsourcing on the Road
Abstract:
Mobile applications are growing fast due to pervasive usage of mobile devices. With inherently limited on-device resources, plenty of research has been conducted on job partitioning/outsourcing strategies to execute mobile computing tasks on external sources, such as public clouds or nearby computers. However, little is known about the performance difference to mobile users on these external computing sources. In this paper, considering the user’s response time and the battery power consumption on mobile devices, we first show that outsourcing mobile applications to public clouds may not outperform outsourcing to nearby residential computers, particularly for delay sensitive applications. To facilitate efficient mobile outsourcing to residential computers, we propose to build a framework RoseMic (ROad-SidE-MobIle-Computing). In RoseMic, a resource overlay network is built with users idle residential (home) computers. To encourage the sharing of idle residential computers,'RoseMic also includes a credit based incentive mechanism that can be enforced automatically without users’ interferences in order to defeat collusion attacks. To demonstrate the performance of RoseMic, we run several real-world applications. The results show that RoseMic outperforms Amazon EC2 by 3 times and 4 times on average in terms of response time and the battery power consumption, respectively.