Optimizing task allocation for edge compute micro-clusters

Alhaizaey, Yousef (2023) Optimizing task allocation for edge compute micro-clusters. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2023AlhaizaeyPhD.pdf] PDF
Download (5MB)


There are over 30 billion devices at the network edge. This is largely driven by the unprecedented growth of the Internet-of-Things (IoT) and 5G technologies. These devices are being used in various applications and technologies, including but not limited to smart city systems, innovative agriculture management systems, and intelligent home systems. Deployment issues like networking and privacy problems dictate that computing should occur close to the data source at or near the network edge.

Edge and fog computing are recent decentralised computing paradigms proposed to augment cloud services by extending computing and storage capabilities to the network’s edge to enable executing computational workloads locally. The benefits can help to solve issues such as reducing the strain on networking backhaul, improving network latency and enhancing application responsiveness. Many edge and fog computing deployment solutions and infrastructures are being employed to deliver cloud resources and services at the edge of the network — for example, cloudless and mobile edge computing.

This thesis focuses on edge micro-cluster platforms for edge computing. Edge computing micro-cluster platforms are small, compact, and decentralised groups of interconnected computing resources located close to the edge of a network. These micro-clusters can typically comprise a variety of heterogeneous but resource-constrained computing resources, such as small compute nodes like Single Board Computers (SBCs), storage devices, and networking equipment deployed in local area networks such as smart home management. The goal of edge computing micro-clusters is to bring computation and data storage closer to IoT devices and sensors to improve the performance and reliability of distributed systems. Resource management and workload allocation represent a substantial challenge for such resource-limited and heterogeneous micro-clusters because of diversity in system architecture. Therefore, task allocation and workload management are complex problems in such micro-clusters.

This thesis investigates the feasibility of edge micro-cluster platforms for edge computation. Specifically, the thesis examines the performance of micro-clusters to execute IoT applications. Furthermore, the thesis involves the evaluation of various optimisation techniques for task allocation and workload management in edge compute micro-cluster platforms. This thesis involves the application of various optimisation techniques, including simple heuristics-based optimisations, mathematical-based optimisation and metaheuristic optimisation techniques, to optimise task allocation problems in reconfigurable edge computing micro-clusters.

The implementation and performance evaluations take place in a configured edge realistic environment using a constructed micro-cluster system comprised of a group of heterogeneous computing nodes and utilising a set of edge-relevant applications benchmark. The research overall characterises and demonstrates a feasible use case for micro-cluster platforms for edge computing environments and provides insight into the performance of various task allocation optimisation techniques for such micro-cluster systems.

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: Singer, Dr. Jeremy
Date of Award: 2023
Depositing User: Theses Team
Unique ID: glathesis:2023-83734
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 24 Jul 2023 10:02
Last Modified: 24 Jul 2023 10:05
Thesis DOI: 10.5525/gla.thesis.83734
URI: https://theses.gla.ac.uk/id/eprint/83734
Related URLs:

Actions (login required)

View Item View Item


Downloads per month over past year