Cloudlet-Assisted Heuristic Offloading for Mobile Interactive Applications

Offloading the computation-intensive task of mobile interactive applications into cloudlets has become a popular way to reduce the response time in mobile cloud computing area. In response to the rapid change of mobile environment, researchers have proposed online offloading decisions for various applications. However, the pruning method they used would perform poorly for the non-linear topology of applications. In this paper, we propose a heuristic to search out the near-optimal offloading scheme for the non-linear topology under the realtime constraint. In order to reduce the searching space, we first leverage Partial Stochastic Path (PSP) analysis to randomly divide the topological graph into a set of linear paths. Then for each linear path, One-climb Policy is applied to obtain the optimal offloading decision as there is at most one-time offloading to the cloudlet if ever. In addition, among the multiple random divisions, the one with the most superior performance is selected as the final division. Lastly, we can get the approximate optimal solution with the minimum completion time in realtime. Simulation results illustrate that our heuristic exhibits an excellent performance for non-linear topology and outperforms the state-of-the-art approaches.