A device-to-device task offloading system for mobile device cloud computing

Abstract

Execution speed bothers application developers for smartphones. Intensive applications like 3D games may suffer from significant delays when CPU is busy. Energy is another concern when smartphones are in low battery level but users need them for urgency use. To ease such pains, traditional cloud-computing framework offload tasks to remote cloud for execution. This framework does not work without Internet access. To deal with such problem, we propose to leverage nearby devices for computation offloading, which is known as mobile device cloud computing. In this thesis, we present Dandelion, an ambient cloud computing system for device to device (D2D) task offloading without the assumption of Internet access. The system includes a programmer friendly framework which provides an open interface to hides all offloading details, a link stability model that evaluates link quality to provide a more reliable network environment, and a runtime task scheduler to make offloading decisions according to runtime environment of CPU idleness and battery level. We implement our system in a testbed with 10 Android phones. Extensive experiments demonstrate that the system achieves up to 6.1X execution speedup and saves up to 5.3X energy consumption.

Date of Award	2014
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology