Recently, organizations worldwide have been embracing intelligent digital technologies in order to boost the quality of living within smart-sustainable environments (e.g., homes, campuses, factories, healthcare, vehicles, cities, etc.). The Internet of Things (IoT) is one of the most promising enabling technologies for deploying these environments by creating a worldwide network of a plethora of interconnected physical objects embedded with electronics, software, sensors, and network connectivity, bringing thereby tremendous opportunities and benefits, including convenience, automation, flexibility, and intelligence. However, this explosive expansion of mobile and sensing devices, cloud services, and video traffic has raised unprecedented challenges for network administrators in advanced solutions development to ensure efficient and reliable management of the underlying constrained infrastructure and the corresponding big data, as innumerable heterogeneous IoT and non-IoT services. On the other hand, the traditional network has limited global visibility of the overall architecture and the corresponding available resources because of the coupled control and data planes paradigm. Software-Defined Networking (SDN) is a promising technology that provides a centralized model with pure software for remote control and dynamic configuration of all heterogeneous network resources and services.
For designing efficient and reliable management of smart-sustainable environments, which is the aim of this thesis, we study the hypothesis of developing generic SDN-based engines for monitoring and optimizing the network underlying infrastructure resources and the massive concurrent heterogeneous flows. To meet this goal, four key issues are required to be addressed in our framework and are summarized as follows: (i) What are the essential particularities of the various smart environments to design generic SDN-based engines? (ii) How to design QoS provisioning aware routing mechanisms? (iii) How to design resource optimization aware routing mechanisms? And (iv) Why/How/Where traffic analysis and characterization (e.g., in terms of QoS) should be performed in fully programmable architectures?
To address the main goal of this thesis and the above research questions, we fix five objectives that enable us to design and experimentally validate new network engines related to traffic analysis, QoS provisioning, resource allocation, and energy consumption. Furthermore, different smart environment use cases (e.g., smart industry and education) have been considered in order to review their requirements and characteristics from various perspectives, thereby validating the proposed engines’ commonness. With each environment, we start by reviewing heterogeneous services and the underlying network architecture, where we particularly focus on the softwarization and programmability technologies. Then, based on each network’s purpose, we design, over centralized control systems, reactive and proactive mathematical models and algorithms using the Lagrangian relaxation theory to manage a set of thousands of heterogeneous flows while fulfilling the required QoS, optimizing available resources, reducing energy consumption, and minimizing the routing cost. On the other hand, we demonstrate how traffic analysis and service classification aspects play a crucial role in appropriate routing algorithm design and network performance. Hence, we focus on network traffic heterogeneity analysis, particularly IoT and non-IoT traffic flow identification. We propose an extensive evaluation framework based on machine and deep learning mechanisms for fine-grained service differentiation while preserving users’ privacy and supporting the evolving nature of network traffic, the high classification accuracy, and the low computational complexity. The results of our study show that the deployment of generic engines over the different smart environments is possible with certain modules (e.g., infrastructure recovery, traffic analysis and characterization, etc.), while some other modules (e.g., QoS provisioning, resource optimization, etc.) need to be adjusted based on the application purpose. Furthermore, the proposed SDN-based engines within each environment significantly enhance QoS assurance, optimize resource allocation, and save energy consumption compared to existing works.
| Date | 10 Dec 2021 |
|---|
| Original language | American English |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Mohamed Cheriet (Supervisor) |
|---|