Internet data traffic will maintain its rapid growth in the next future due to the ever increasing development of new application services and devices. The 4G network is not able to meet this explosive growth in traffic demand. For this reason, 5G mobile network represents an optimal solution due to its designed architecture, modifying the actual ossified infrastructure thanks to the use of paradigms like NFV, SDN and MEC. A key feature of 5G is Network Slicing, a technology that allows the creation of virtual networks able to guarantee different applications requirements. In this context, this thesis was realized with the aim of studying the main aspects of network softwarization in 5G environment, from the management and orchestration of network slices to their use for vertical services. In particular, as regards the management and orchestration, a first work concerns the definition of a framework for the prediction of handovers, aimed at optimizing resources in scenarios characterized by high variability of the access point by the users. In the same context, the use of UAVs to extend network slices is proposed, providing networking and computing resources at the edge of the network. Using Reinforcement Learning, it is possible to offload part of the data to be processed from one UAV to another nearby, optimizing parameters such as latency, loss probability and energy consumption. About the use of network slices for vertical services, two works are presented: the first concerns the definition of the functional architecture of the Tactile Support Engine in the Tactile Internet network slice. The second proposed work concerns hierarchical offloading in vehicle networks based on Reinforcement Learning techniques. Finally, thanks to the participation in European projects, it was possible to define and implement some application scenarios characterizing a 5G ecosystem, taking care of three areas that are now considered strategic: distributed video surveillance, development of assistive technologies for patients suffering from various kinds of diseases and online gaming.
Il traffico dati su Internet manterrà la sua rapida crescita nel prossimo futuro grazie allo sviluppo sempre maggiore di nuovi dispositivi servizi applicativi. La rete 4G non è in grado di soddisfare questa crescita esplosiva della domanda di traffico. Per questo motivo, la rete mobile 5G rappresenta una soluzione ottimale grazie alla sua architettura, modificando l'attuale infrastruttura ossificata grazie all'utilizzo di paradigmi come NFV, SDN e MEC. Una caratteristica fondamentale del 5G è il Network Slicing, una tecnologia che permette la creazione di reti virtuali in grado di garantire diverse esigenze applicative. In questo contesto, questa tesi è stata realizzata con l'obiettivo di studiare i principali aspetti della softwarizzazione della rete in ambiente 5G, dalla gestione e orchestrazione delle network slice al loro utilizzo per servizi verticali. In particolare, in merito alla gestione e orchestrazione, un primo lavoro riguarda la definizione di un framework per la previsione degli handover, finalizzato all'ottimizzazione delle risorse in scenari caratterizzati da elevata variabilità del punto di accesso da parte degli utenti. Nello stesso contesto, viene proposto l'uso di UAV per estendere le network slice, fornendo risorse di rete e di computing ai margini della rete. Utilizzando il Reinforcement Learning, è possibile eseguire l'offloading di parte dei dati da elaborare da un UAV a un altro nelle vicinanze, ottimizzando parametri come latenza, probabilità di perdita e consumo energetico. Sull'utilizzo di network slice per servizi verticali, vengono presentati due lavori: il primo riguarda la definizione dell'architettura funzionale del Tactile Support Engine nella network slice Tactile Internet. Il secondo lavoro proposto riguarda l'offloading gerarchico nelle reti veicolari basate su tecniche di Reinforcement Learning. Infine, grazie alla partecipazione a progetti europei, è stato possibile definire e implementare alcuni scenari applicativi caratterizzanti un ecosistema 5G, occupandosi di tre ambiti ormai considerati strategici: la videosorveglianza distribuita, lo sviluppo di tecnologie assistive per pazienti affetti da disabilità di varia natura e il gaming online.
Softwarizzazione e servizi smart in sistemi di rete 5G / Grasso, Christian. - (2021 Feb 02).
Softwarizzazione e servizi smart in sistemi di rete 5G
GRASSO, CHRISTIAN
2021-02-02
Abstract
Internet data traffic will maintain its rapid growth in the next future due to the ever increasing development of new application services and devices. The 4G network is not able to meet this explosive growth in traffic demand. For this reason, 5G mobile network represents an optimal solution due to its designed architecture, modifying the actual ossified infrastructure thanks to the use of paradigms like NFV, SDN and MEC. A key feature of 5G is Network Slicing, a technology that allows the creation of virtual networks able to guarantee different applications requirements. In this context, this thesis was realized with the aim of studying the main aspects of network softwarization in 5G environment, from the management and orchestration of network slices to their use for vertical services. In particular, as regards the management and orchestration, a first work concerns the definition of a framework for the prediction of handovers, aimed at optimizing resources in scenarios characterized by high variability of the access point by the users. In the same context, the use of UAVs to extend network slices is proposed, providing networking and computing resources at the edge of the network. Using Reinforcement Learning, it is possible to offload part of the data to be processed from one UAV to another nearby, optimizing parameters such as latency, loss probability and energy consumption. About the use of network slices for vertical services, two works are presented: the first concerns the definition of the functional architecture of the Tactile Support Engine in the Tactile Internet network slice. The second proposed work concerns hierarchical offloading in vehicle networks based on Reinforcement Learning techniques. Finally, thanks to the participation in European projects, it was possible to define and implement some application scenarios characterizing a 5G ecosystem, taking care of three areas that are now considered strategic: distributed video surveillance, development of assistive technologies for patients suffering from various kinds of diseases and online gaming.File | Dimensione | Formato | |
---|---|---|---|
Tesi di dottorato - GRASSO CHRISTIAN 20201115174259.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
PUBBLICO - Pubblico con Copyright
Dimensione
17.68 MB
Formato
Adobe PDF
|
17.68 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.