11th International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services

Research Article

Optimal Predictive Code Offloading

  • @INPROCEEDINGS{10.4108/icst.mobiquitous.2014.258023,
        author={Florian Berg and Frank D\'{y}rr and Kurt Rothermel},
        title={Optimal Predictive Code Offloading},
        proceedings={11th International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services},
        keywords={preemptable code offloading connectivity prediction user movement mobile cloud computing},
  • Florian Berg
    Frank Dürr
    Kurt Rothermel
    Year: 2014
    Optimal Predictive Code Offloading
    DOI: 10.4108/icst.mobiquitous.2014.258023
Florian Berg1,*, Frank Dürr1, Kurt Rothermel1
  • 1: Institute of Parallel and Distributed Systems, University of Stuttgart
*Contact email: Florian.Berg@ipvs.uni-stuttgart.de


Modern mobile devices like smart phones and tablets are equipped with powerful processing and memory resources, enabling resource-intensive mobile applications such as high-end mobile games. The main limitation, however, remains the energy resource. To improve the energy efficiency, code offloading has been proposed, which offloads code to remote servers and transfers the results back to the mobile device. Although several approaches have shown that code offloading improves energy efficiency significantly in general, they largely neglect the adverse effects of network disconnections. Therefore, we have proposed the concept of preemptive code offloading to improve energy efficiency also under link failures. It transmits so-called safe-points between server and mobile device during remote execution, enabling the re-use of partial remote results after link failures. In this paper, we improve our basic preemptive code offloading approach by optimizing the time when to generate and transmit safe-points to minimize the communication overhead and maximize energy efficiency. To find the optimal safe-point schedule, we use a predictive approach that predicts the mobile link quality in order to send safe-points before network disconnections. Moreover, we consider additional deadline constraints for code execution to ensure a minimal responsiveness of offloaded applications despite link failures. Our evaluation results show that energy efficiency can be improved significantly using our predictive offloading approach.