QoE-sensitive softwarized networks using AI-based inference

Publié le
Equipe
Date de début de thèse (si connue)
dès que possible
Lieu
IRISA
Unité de recherche
IRISA - UMR 6074
Description du sujet de la thèse

1. Context

Network softwarization has been proposed as an alternative way for computer networking. This environment has opened up a new era in computer networks where innovative solutions are sought. Undoubtedly, network softwarization is a primary driver behind the thriving 5G/6G and Edge/Core systems. In a software network environment, there is a sharp distinction between pictured and physical networks. A pictured network is an instantiation of a real network that exists for a certain period of time over a physical network. A pictured network, which is ad—hoc and volatile, exploits physical resources provided by the physical network. It utilizes tailored transport functions and policies in order to manage and control assured traffic and services.

2. Motivation

In the context of network softwarization, innovative approaches and solutions should be proposed for the sake of networks management, control and monitoring of pictured networks. It is highly desirable to consider QoE scores when executing these network operations. This enables to undertake actions that achieve exactly what is required by end-users. However, the measurement of QoE over software network is still in its infancy and needs an extensive exploration and novel instrumentations. Moreover, considering QoE measurement over a software network is an open issue that requires a deep investigation from theoretical and practical perspectives.

3. Objectives

This research work aims at providing new solutions and methods in order to enable QoE-aware management of software networks. The first step consists of proposing an architecture for monitoring QoE accounting for softwarized network environment. It should support different kind of services that could be provided over software networks, e.g. VoIP and video streaming. The proposed architecture should incorporate a protocol enabling quick and automatic evolution of the measurement processes and their configuration. The second step consists of providing an optimized instrumentation strategy, which is automatically specified as a function of parameters required by each invoked QoE function, and the enabled actions by the data plane. The third and final step consists of proposing advanced estimators of missing or inaccessible parameters required by a given QoE function. This relies on AI-based inference techniques that process a reduced and easily-accessible dataset in order to extract valuable and insightful metrics leading to an accurate QoE estimation.

 

 

Bibliographie

[1] Z. Yang and K. L. Yeung, "Flow monitoring scheme design in SDN", Computer Networks, Volume 167, 2020.

[2] l. A. Barakabitze and R. Walshe, "SDN and NFV for QoE-driven multimedia services delivery: The road towards 6G and beyond networks", Computer Networks, Volume 214,
2022,

[3] A. Wang, Y. Guo, F. Hao, T. Lakshman and S. Chen, "UMON: Flexible and Fine Grained Traffic Monitoring in Open vSwitch", ACM CoNEXT, 2015.

[4] Z. Zha, A. Wang, Y. Guo, D. Montgomery and S. Chen, "Instrumenting Open vSwitch with
Monitoring Capabilities: Design and Challenges", ACM SOSR 2018.

[5] L. Yu, Z. Li, Y. Zhong, Z. Ji and J. Liu, "When QoE meets learning: A distributed traffic-processing framework for elastic resource provisioning in HetNets", Computer Networks, Volume 167, 2020.

[6] R. Nikbazm and M. Ahmadi, "KSN: Modeling and simulation of knowledge using machine learning in NFV/SDN-based networks", Simulation Modelling Practice and Theory, Volume 121, 2022.

Liste des encadrants et encadrantes de thèse

Nom, Prénom
Jelassi, Sofiene
Type d'encadrement
Co-encadrant.e
Unité de recherche
IRISA
Equipe

Nom, Prénom
Sericola, Bruno
Type d'encadrement
Directeur.trice de thèse
Unité de recherche
IRISA
Equipe
Contact·s
Nom
Jelassi, Sofiene
Email
sofiene.jelassi@irisa.fr
Téléphone
0767663843
Mots-clés
Network softwarization, AI-based network metrology, QoE monitoring, metrics' inference