Enhancing IoT services in mobile edge computing: offloading strategies and SFC placement techniques
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
With the rapid proliferation of mobile devices, applications such as face recognition, online gaming, and video streaming are becoming prevalent for their capacity to engage users. These applications are typically resource-hungry, exceeding the computational and energy capacities of mobile devices, often leading to unsatisfactory user experience. Traditional cloud computing faces challenges in meeting these applications requirements resulting in significantly increased latency and network congestion. Mobile Edge Computing (MEC) has emerged as a solution by allowing computational tasks to be offloaded to physically proximate mini-datacenters, called cloudlets. This is further enhanced by integrating Network Function Virtualization (NFV) into MEC, accelerating service placement to deliver more efficient network services and offloading. While MEC and NFV present numerous opportunities to enhance both the Quality of Experience (QoE) for users and the Quality of Service (QoS) for the overall network, they also introduce unique challenges due to the constrained coverage and limited computational resources of cloudlets, as well as complexities of managing dependencies among network functions. ☐ This dissertation aims to address key challenges in computation offloading and Service Function Chains (SFCs) placement. Firstly, we study the task offloading problem among a swarm of capacitated Unmanned aerial vehicles (UAVs) acting as flying cloudlets. We devise a novel capacitated offloading game that confirms the existence of a Nash equilibrium, considering the incentives of UAVs. Secondly, we study the offloading problem from a vehicle to a group of vehicles with available computational resources via vehicle-to-vehicle (V2V) communications, considering data privacy and accessibility. We propose privacy-by-design offloading solutions to facilitate latency requirements and reduce energy consumption of vehicles. Thirdly, we explore the truthfulness of users’ preferences when they request edge services. We design a novel incentive-compatible offloading mechanism to implement an efficient system equilibrium. Fourthly, we define a video offloading problem in MEC considering uncertainty caused by device mobility. We devise two efficient uncertainty-aware approaches that do not have any information about users future mobility, and each resonating with a unique perspective: system-side and user-side. Last but not least, we investigate the multi-SFC placement problem in MEC-NFV networks to jointly minimize the deploying cost of placing VNFs on cloudlets and the routing cost of transferring outputs from predecessor functions. We propose a novel approach that converts SFCs into a directed acyclic graph, followed by an innovative algorithm based on topological sort. ☐ The complexities inherent in MEC are due to several key factors. First, cloudlets in MEC have varying and limited storage and computing capacities, which significantly challenges the optimal placement of computational tasks and complex SFCs among these cloudlets. Additionally, the dependencies among VNFs in a chain add complexity to the SFC placement problem. Moreover, the diverse preferences of users, such as the expectation of low latency, minimal energy consumption, and strict privacy requirements, further complicate the offloading problem. Another significant challenge is user mobility. The dynamic changes in user positions introduces uncertainties that exacerbate the complexity of these problems. To address these challenges, this dissertation contributes scalable solutions to these multifaceted problems and outlines future research directions that may stem from this research.
Description
Keywords
Internet of Things, Mobile Edge Computing, Network Function Virtualization, Quality of Experience, Unmanned aerial vehicles, Energy consumption