AbstractIn recent years, there has been considerable interest in computational offloading from mobile devices (such as smartphones) to reduce computational task completion time and the associated energy consumption by battery-dependent mobile devices.
The most researched strategy involves offloading to Cloud Computing in Mobile Cloud Computing, where the Cloud offers not only storage but Software as a Service. Resorting to physically distant consolidated data centres, however, has excessive latency and low bandwidth issues.
Multi-access Edge Computing (MEC) networks offer improved solutions to real-time and delay-sensitive mobile applications with closer proximity of cellular networks and potentially much larger numbers of hardware units accessible by mobile devices.
This thesis explores different scenarios for offloading computational tasks to MEC servers from multiple users with a range of mobile devices. A unified and coherent approach presents detailed simulation data for how offloading can be beneficial to reduce total task completion time and local (mobile device) energy use in MEC networks with varying quantitative mobile user demand, heterogeneity in mobile device on-board and MEC processor speeds, communication speeds, link access delays and mobile device numbers. The analysis is then extended to show that the relationship between CPU workloads on the mobile device and a MEC server and the link speed between them are crucial parameters determining the success of offloading to the MEC network to reduce total task completion time and mobile device energy use.
Furthermore, novel distributed heuristic algorithms have been developed that allow a mobile device to decide how to select the least-time schedules of multiple jobs to be offloaded and to identify least-time solutions for multiple mobile devices simultaneously offloading jobs to multiple MEC servers. The proposed heuristic algorithms are tested in a range of numerical simulations, and the results demonstrate that the heuristic approach can produce reasonable quality solutions in comparison with linear programming.
Heuristic algorithms have been extended to incorporate time and energy in a network with multiple MEC servers and mobile devices MDs to focus on an objective function with variable weighting factors for time and local energy use; this approach is designed to give the use of a mobile device the maximum flexibility in choosing savings for time and energy use. Numerical simulations in test cases evaluate the impact of changing weighting factors. The objective function shows a continuous variation as the emphasis is placed on either time or energy saving by the weighting factors. The numerical tests also demonstrate that the proposed heuristic algorithms produce near-optimal computational offloading solutions using a combined weighted score for schedule task completion time and energy.
A preliminary multi-factorial analysis includes economic cost factors to explore how a subscription service could reflect mobile device users’ varying requirements in minimising task completion time or extending the battery lifetime of mobile devices..
|Date of Award||28 Sep 2021|
|Supervisor||Simon M D Armour (Supervisor) & George Oikonomou (Supervisor)|