Democritus University of Thrace (DUTH)
Μεθοδολογίες βέλτιστης διαχείρισης δεδομένων και πόρων στο Edge Computing
Abstract
dc:descriptionIn recent years, due to the evolution of Edge Devices and networks, a remarkable increase in the use of Edge Computing has been sighted in many topics, such as medicine and industry. Edge Computing solves many problems, such as real-time data processing with reduced latency and low power consumption of devices. This is where the need and importance of optimal data and resource management methodologies are presented, topics that are attracting increasing interest from the scientific community. This thesis focuses on using an approach that departs from the traditional offloading approach, in which Edge Devices offload their tasks to Servers or even to the Cloud. More specifically, we enhance a scenario where Edge Devices assist the more powerful nodes in executing processes. As the number of Edge Devices increases, an abundance of idle computing capacity is created, which, when aggregated in a systematic fashion, can be proven beneficial. We adopt this idea of reverse offloading and study a scenario where a powerful node distributes processes to a set of nearby Edge Devices. Our main contribution consists in the implementation of two innovative methodologies with the logic of reverse task offloading. The first methodology solves a single-objective optimization problem for the optimal distribution of Artificial Intelligence processes as well as the optimal choice of the AI model for each Edge Device, aiming to optimize the processing time, taking into account execution accuracy and energy consumption constraints. The second methodology extends the above logic by solving a multi-objective optimization problem, targeting the simultaneous optimization of three objectives: the AI task execution time, the execution accuracy and the overall energy consumption of the Edge Computing system. The results of this thesis show that the proposed methodologies improve the system performance compared to baseline algorithms in the literature.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Amanatidis, Petros
- Αμανατίδης, Πέτρος
Subjects
dc:subject × 14- Εκφόρτωση διεργασιών
- Μονοκριτηριακή βελτιστοποίηση
- Πολυκριτηριακή βελτιστοποίηση
- Ανίχνευση αντικειμένων
- Task offloading
- Single objective optimization
- Multi objective optimization
- Object detection
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Επιστήμη ηλεκτρονικών υπολογιστών και Πληροφορική, άλλοι τομείς
- Computer and Information Sciences
- Natural Sciences
- Computer and Information sciences, miscellaneous
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59333