Computation offloading in mobile edge computing: an optimal stopping theory approach

Alghamdi, Ibrahim (2021) Computation offloading in mobile edge computing: an optimal stopping theory approach. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2021AlghamdiIbPhD.pdf] PDF
Download (6MB)

Abstract

In recent years, new mobile devices and applications with different functionalities and uses, such as drones, Autonomous Vehicles (AV) and highly advanced smartphones have emerged. Such devices are now able to launch applications such as augmented and virtual reality, intensive contextual data processing, intelligent vehicle control, traffic management, data mining and interactive applications. Although these mobile nodes have the computing and communication capabilities to run such applications, they remain unable to efficiently handle them mainly due to the significant processing required over relatively short timescales. Additionally, they consume a considerable amount of battery power. Such limitations have motivated the idea of computation offloading where computing tasks are sent to the Cloud instead of executing it locally at the mobile node. The technical concept of this idea is referred to as Mobile Cloud Computing (MCC). However, using the Cloud for computational task offloading of mobile applications introduces a significant latency and adds additional load to the radio and backhaul of the mobile networks. To cope with these challenges, the Cloud’s resources are being deployed near to the users at the Edge of the network in places such as mobile networks at the Base Station (BS), or indoor locations such as Wi-Fi and 3G/4G access points. This architecture is referred to as Mobile Edge Computing or Multi-access Edge Computing (MEC). Computation offloading in such a setting faces the challenge of deciding which time and server to offload computational tasks to.

This dissertation aims at designing time-optimised task offloading decision-making algorithms in MEC environments. This will be done to find the optimal time for task offloading. The random variables that can influence the expected processing time at the MEC server are investigated using various probability distributions and representations. In the context being assessed, while the mobile node is sequentially roaming (connecting) through a set of MEC servers, it has to locally and autonomously decide which server should be used for offloading in order to perform the computing task. To deal with this sequential problem, the considered offloading decision-making is modelled as an optimal stopping time problem adopting the principles of Optimal Stopping Theory (OST).

Three assessment approaches including simulation approach, real data sets and an actual implementation in real devices, are used to evaluate the performance of the models. The results indicate that OST-based offloading strategies can play an important role in optimising the task offloading decision. In particular, in the simulation approach, the average processing time achieved by the proposed models are higher than the Optimal by only 10%. In the real data set, the models are still near optimal with only 25% difference compared to the Optimal while in the real implementation, the models, most of the time, select the Optimal node for processing the task. Furthermore, the presented algorithms are lightweight, local and can hence be implemented on mobile nodes (for instance, vehicles or smart phones).

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Colleges/Schools: College of Science and Engineering > School of Computing Science
Supervisor's Name: Pezaros, Professor Dimitrios P. and Anagnostopoulos, Dr. Christos
Date of Award: 2021
Depositing User: Theses Team
Unique ID: glathesis:2021-82506
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 12 Oct 2021 09:58
Last Modified: 08 Apr 2022 17:07
Thesis DOI: 10.5525/gla.thesis.82506
URI: https://theses.gla.ac.uk/id/eprint/82506
Related URLs:

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year