Espace membre

Cet espace est dédié aux étudiants, aux enseignants et au personnel administratif de l'école


Mot de passe oublié?



Doctorate thesis defense of Amel KSENTINI

Doctorate thesis defense on September 27th 2021 at 10H30 ,in Sup’Com Amphitheater Ibn Khaldoun.

Entitled :Quality of Service management in a Fog Computing environment

Presented by :Amel KSENTINI


President :

Zied Choukair

Professor at SUP'COM, Tunisia




Reviewers :

Lamia Chaari Fourati

Professor at ISIMS, Tunisia


Tijani Chahed

Professor at Telecom SudParis, France


Examiner :

Sihem Guemara

Professor at SUP'COM, Tunisia




Thesis Supervisor :

Sami Tabbane

Professor at SUP'COM, Tunisia


During latest years, Fog Computing has been developed as a novel paradigm that stands out for its closeness to the edge of the network. It is a concept that carries services, habitually held in the Cloud, (compute, storage, processing, etc.) to end users. As a result, Fog Computing is likely to succeed to enhance performances such as reducing latency and guaranteeing better availability of services. Therefore, it is highly recommended and used in real-time use-cases requiring low processing and response time. However, researchers still have to deal with several issues namely related to quality of service (QoS) and energetic aspects. On one hand, regarding QoS management, applications deployed in Fog environment are very multiple and varied. In addition, the deployment of applications has to deal with a heterogeneous and very dynamic environment. On the other hand, in such environment, a very large number of service requests, coming simultaneously from sensors and end users devices, are managed. Therefore, energy optimization is paramount. It is therefore recommended to deal with the problem of energy management in a Fog Computing system by proposing solutions that take into account the specificities of Fog nodes. Thus, the objective of this thesis is to deal with both energy and quality aspects of a Fog Computing system. The investigation of existing architectures enabled us to deduce a reference model of the Fog Computing system for the management of QoS. Moreover, the study of energy management algorithms helped us to focus on improving the QoS while ensuring efficient energy consumption. Thereby, we introduced a reference model and a potential solution for QoS management inspired from the architectures survey. By exploiting our proposed reference model for fog environment, we focused on the benefit of fog nodes, implied in IoT/Cloud integration architecture. Thus, we investigated on how Fog computing paradigm can help reducing energy consumption for delay-sensitive services in smart cities, while maintaining several required QoS metrics. We conclude that we can reduce latency in a delay-sensitive application without increasing the power consumption, even better, the energy consumption is decreased as well as the network usage of each application and so, of the overall system. The next step was to conduct a review on QoS requirements in fog environment, which enabled us to come out with a proposed approach based on priority classification (PC) technique. This contribution aims to achieve an efficient deployment of smart application in Fog computing systems, on the basis of a classification priority approach, mapped into hierarchy models for each class of priority. Although the novel model achieves better performance compared to a Cloud-only model, the distributed differentiation process at the fog nodes also plays a considerable role in the deteriorating of several QoS requirements such as increasing processing time and latency, while increasing energy consumption of the overall system. Centralizing the PC process over a distinguished control plan was a suitable solution to solve the abovementioned issues. Therefore, we considered an industrial IoT deployment enabled by Fog computing, focusing on the QoS management. We presented a novel slicing model based on a PC scheme. We used a multi-objective optimization algorithm involving data-rate, latency, cost and energy consumption metrics to set the slicing strategy, and performed with a Machine Learning workbench. The proposed network slicing model achieves better performances than a reference model, in terms of data-rate, e2e delay and network usage.

Keywords :

Fog Computing, QoS, IoT, Cloud Computing, SDN, NFV, Machine Learning, Slicing, Energy Consumption.