Description
Date depot: 23 mars 2022
Titre: Virtual Network Functions Management with Machine Learning, for beyond 5G networks
Directeur de thèse:
Serge FDIDA (LIP6)
Encadrant :
Thanasis KORAKIS (CERTH)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Systèmes et réseaux
Resumé: Wireless communication systems play a very critical role in modern society for recreational, business, commercial, health and safety applications. 5th generation systems (5G) is already a reality, offering unprecedented speeds and ubiquitous connectivity. Industry and universities are active around the evolution of such systems beyond 5G and 6G systems, which will offer higher speeds, will support more connected customers, and will facilitate largely the management of distributed infrastructure. 5G and its successors are going to have a dependence on software compared to previous generations, based on wide distribution used throughout the stack and disconnection of the different functions from the specific hardware of the closed source vendor. This wide application software leads to further innovations by executing the various parts of the stack as separate functions in the cloud, allowing stack virtualization. This virtual software is the main assistant for the dynamic sharing of its resources in a number of sections. Each slice can be created from end to end, offering guarantees for specific services and a set of users, including one Virtual Network Function (VNF) in conjunction with other network resources (e.g. capacity at base stations). Within non-5G networks, this calculation expands to the base station stack, thus allowing integration of more functions in the VNF chain. However, as these functions were hosted previously on exclusive material, special care must be taken when performing them as VNF, in relation to parameters such as network load, service level agreements (SLAs) provided through the network and the placement of network functions.
Artificial Intelligence (AI) can play an important role in this context, applying Machine Learning methods to predict the evolution of specific network parameters, resulting in appropriate decisions that can meet the requirements of the network (e.g. network load, SLA retention, reduction of radio access delay network). Such decisions may include determining the scale of the network, the relocation of the services closer to the edge of the network or the installation of new features within the network. Developments in artificial intelligence can help improve efficiency and effectiveness of the 5G system, while their multiplication of the 5G connected devices can lead to distributed intelligence with continuous improvements in learning and AI conclusions.
The primary goal of AI integration is to optimize network performance and improve QoS (Quality of Service) and QoE (User Experience Quality) by anticipating the upcoming load of resources and making smarter decisions faster and more efficiently. In this doctoral dissertation, all these open issues will be studied extensively, while emphasis will be placed on orchestrating Virtual Network Functions (VNFs) as light micro-services with the help of Machine Learning algorithms that evaluate each problem, make predictions and take the appropriate actions in an experimental set that represents real conditions. Specifically, the objectives of the doctorate are:
• The study, design and experimental evaluation of new algorithms for the placement of VNFs in a Telecommunications Network based on available computing resources.
• The study and design of various machine learning algorithms for a more efficient implementation of a VNF Telecommunication Network in cloud environment.
• The study and design of scaling algorithms in a cloud environment using Artificial Intelligence for predicting the upcoming load and making the right decisions for consuming VNF resources (CPU, Memory, service playback etc.) dynamically.
The following diagram will be followed for its elaboration:
• Study of the relevant literature.
• Design of a modern telecommunications network beyond 5G.
• Application of new algorithms for estimation and study of network conditions and characteristics of its available resources.
• Merge machine learning algorithms with cloud computing for making real-time decisions based on trends and characteristics of the past.
• Evaluation of proposed architectures and algorithms through extensive experiments in a simulated real-life environment.
• Thesis writing and support.
References:
1. J. Kaur, M. A. Khan, M. Iftikhar, M. Imran and Q. Emad Ul Haq, "Machine Learning Techniques for 5G and Beyond," in IEEE Access, vol. 9, pp. 23472-23488, 2021, doi: 10.1109/ACCESS.2021.3051557.
2. T. Wang, S. Wang and Z. -H. Zhou, "Machine learning for 5G and beyond: From model-based to data-driven mobile wireless networks," in China Communications, vol. 16, no. 1, pp. 165-175, Jan. 2019, doi: 10.12676/j.cc.2019.01.015.
3. M. M. da Silva, & J. Guerreiro. (2020). On the 5G and Beyond. Journal of Applied Sciences (Switzerland), MDPI, 10(20), 1-12.
Doctorant.e: Zalokostas-Diplas Vasileios