A resource management framework for fog computing networks
| dc.contributor.author | Mtshali, Mxolisi | |
| dc.date.accessioned | 2023-03-14T10:31:08Z | |
| dc.date.available | 2023-03-14T10:31:08Z | |
| dc.date.issued | 2020 | |
| dc.description | A dissertation submitted in fulfillment of the academic requirements for the degree of Master of Science in the Department of Computer Science in the Faculty of Science, Agriculture and Engineering, University of Zululand, 2020. | en_US |
| dc.description.abstract | The evolution of the Internet of Things (IoT) has drastically changed how computing devices can be deployed by enabling them to be located anywhere on the cloud-to-things continuum. This has emerged as a solution to most emerging market challenges. These ubiquitous computing devices do not only possess the data-processing capabilities, they also have computing and storage capabilities. Based on these device capabilities, a new paradigm has evolved. This paradigm decentralizes all the centralized cloud capabilities and locates them at the edge of the network. This is popularly known as fog computing. The goal is to avoid deploying the IoT services to the cloud core servers for processing and storage resources, to also mitigate latency, and cost of deployment. Since the paradigm is relatively new, there are challenges that need to be tackled in order to have a reliable network deployment. The main challenge of this paradigm arises from the relocation of service from the resource-rich physical underlying infrastructure of the cloud core to the Fog layer where there is limited resources and physical infrastructure, because limited infrastructure offers limited capacity in terms of computing, network, processing and storage resources. This means that should a heavy application be executed on these limited FN services, they can be over-consumed, which could lead to network breakdown or failure owing to device shutdown. As the resources are the main significant part of the network, they must be able to offer services on demand without being exploited. Therefore, in order to optimize the resource management operation efficiently, a method such as resource scheduling must be applied to fog deployments. To address the matter, factors such as scheduling algorithms and framework are considered. In the process of network design and deployment, critical questions arise, such as: How can a resource management framework be used to address the challenges in fog computing; Why do existing resource scheduling mechanisms not respond adequately to resource management challenges of fog computing networks? Which resource-scheduling algorithms can be used to address specific resource management challenges, and what are their relevant achievements? Henceforth, this challenge will be referred to as a resource management problem. This challenge is a multi- objective problem and comprises competing objectives such as service delay, energy consumption and network utilization. Therefore, it affirms that there is no universal solution available. Therefore, this dissertation proposes a resource management framework as a solution to the network-planning problem. The planning covers the optimal placement of tasks with respect to resource consumption optimization and optimal offloading of tasks in the continuum. Addressing this multi-objective optimization problem involves two stages. First, four classical scheduling methods, namely, first come first serve (FCFS), shortest job first (SJF), round robin (RR), and priority-based (PB) have been evaluated. The optimal method progresses from the first stage to the second stage, where its overall performance is evaluated against an unsupervised machine-learning algorithm K-means. The simulation findings show that, as the number of sensor request scales up, the service delay and the energy consumption of the FCFS scheduling method increase linearly; while in the case of K-means, the service delay increases exponentially. The FCFS method yields optimal results in terms of service delay and CPU execution time spent on the task, and also achieved the best trade-off between the competing objectives of service delay and energy consumption, whereas, the K-means clustering method has optimal energy consumption with high service delay resulting in the worst trade-off. The modelling covers most realistic fog computing applications, parameters and constraints, therefore it can be easily deployed in the fog computing landscape while extending the current cloud architecture. | en_US |
| dc.identifier.uri | https://hdl.handle.net/10530/2277 | |
| dc.language.iso | en | en_US |
| dc.publisher | University of Zululand | en_US |
| dc.subject | resource management framework | en_US |
| dc.subject | fog computing networks | en_US |
| dc.title | A resource management framework for fog computing networks | en_US |
| dc.type | Thesis | en_US |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- A resource management framework for fog computing networks.pdf
- Size:
- 1.41 MB
- Format:
- Adobe Portable Document Format
- Description:
- A resource management framework for fog computing networks
License bundle
1 - 1 of 1