Mobility Prediction Based Opportunistic Computational Offloading for Mobile Device Cloud

Abstract

In mobile cloud computing environments, it's regarded as a good solution to augment the capability of the resource-constrained devices by offloading some of their computation-intensive applications to other more powerful surrogate devices to execute. However, because the nodes are usually connected via certain wireless technology and the nodes may change their locations from time to time, the connections between devices are usually unstable and the applications offloaded may fail. In order to guarantee the users to be able to continue the applications offloaded seamlessly regardless of the mobility of the nodes, in this paper, the extended versions of the traditional Minimum Execution Time heuristic and the Minimum Completion Time heuristic, and a mobility prediction based offloading heuristic, were proposed to solve the mobility problem in mobile device clouds. Their performances were investigated via simulation. It's shown that, with the help of mobility prediction, the Dyn Predict heuristic can lead to lower average reschedule time, lower average failure rate and shorter response time.