Internet of Things (IoT) devices are the key enabler for the envisioned smart cities. In many application scenarios, 5G and Edge computing will work together to bring computational power and storage closer to IoT devices. Nevertheless, cloud-based IoT systems do not make use of edge computing which can grant real-time communication between devices, greater data privacy and location awareness support. In addition, many communication protocols still rely on centralised systems instead of the envisioned distributed architecture at the network’s edge. In this Ph.D. thesis, we study two aspects of the convergence between IoT and 5G edge computing. First, designing and implementing two IoT applications that offload computational-intensive tasks to the edge. Second, proposing a distributed publish-subscribe communication protocol for IoT edge computing nodes. Special attention is paid to the MQTT pub/sub protocol and the role of communication between MQTT brokers. In our first contribution, we propose a combined solution of two IoT prototypes for a smart campus scenario, called Smart Waste Bin and Smart Gate. To support complex decision algorithms and information fusing, the device intelligence runs on the edge of the 5G network rather than on a cloud server or locally on the devices themselves. We discuss the entire design of the system prototypes, from the analysis of requirements to the implementation details. Results indicate that moving the artificial intelligence to the edge of the network is beneficial from latency and energy consumption perspectives. The second contribution presents a new benchmarking framework for distributed MQTT brokers. The system, called BORDER, helps to evaluate the general performance of MQTT brokers beyond the envisioned application domain. Finally, in our main contribution, we propose an evolution of the MQTT protocol. Currently, it is based on a single-server topology that does not scale well considering the massive numbers of IoT devices expected in the next future. In this thesis, we aim to implement a distributed MQTT broker system to increase failure recovery, network scalability, and, thus, make MQTT even more effective in the IoT and edge domain. We explore three different algorithms to achieve such a distribution, focusing particularly on the creation of the MQTT brokers overlay network. The stages of the MQTT evolution are tested in different scenarios and performance metrics such as end-to-end delay, resource and bandwidth consumption are analysed. In conclusion, this work shows the capabilities of edge-enabled IoT devices and a solution for efficient communication systems for distributed servers located at the edge of the network.
I dispositivi IoT (Internet of Things) sono l'elemento chiave per le città intelligenti del futuro. In diversi use-case, il 5G e l'Edge computing lavoreranno insieme per portare potenza di calcolo e spazio di archiviazione più vicina ai dispositivi IoT. Tuttavia, la maggioranza dei sistemi IoT sono basati sul cloud, quindi non utilizzando l'edge computing, che faciliterebbe una comunicazione real-time, maggior privacy dei dati e mobilità dei dispositivi. Inoltre, anche i protocolli di comunicazione IoT si basano ancora su sistemi centralizzati invece che sull'architettura distribuita all’edge della rete. In questa tesi di dottorato, abbiamo studiato due aspetti della convergenza tra IoT e 5G edge computing. In primis, sono stati progettati e implementati due applicazioni IoT che delegano le operazioni ad alta intensità di calcolo all'edge. Successivamente, è stato proposto un protocollo di comunicazione distribuito publish-subscribe per i nodi IoT all’edge. Particolare attenzione è stata rivolta al protocollo MQTT pub/sub e al ruolo della comunicazione tra i broker MQTT. Nel primo contributo, presentiamo una soluzione congiunta di due prototipi IoT per uno scenario di smart campus, chiamati Smart Waste Bin e Smart Gate. Per supportare algoritmi decisionali complessi e l’information fusion, l'intelligenza del dispositivo viene spostata ai margini della rete 5G piuttosto che su un server cloud o sul dispositivo stesso. Durante la tesi abbiamo illustrato l'intera progettazione dei prototipi, dall'analisi dei requisiti ai dettagli dell’implementazione. I risultati mostrano che l’utilizzo dell'intelligenza artificiale all’edge della rete porta notevoli vantaggi dal punto di vista della latenza e del consumo energetico. Il secondo contributo presenta un nuovo framework per il benchmark di broker MQTT distribuiti. Il sistema implementato, chiamato BORDER, aiuta a valutare le prestazioni dei broker MQTT per le applicazioni IoT. Infine, nel contributo principale, proponiamo un'evoluzione del protocollo MQTT. Attualmente, il protocollo è basato su una topologia a single-server che non scala se si considera l'enorme numero di dispositivi IoT previsti in futuro. In questa tesi, mostriamo lo studio e l’implementazione di un sistema di broker MQTT distribuito per gestire il ripristino dopo una failure, aumentare la scalabilità della rete e, quindi, rendere MQTT ancora più efficace nel dominio IoT ed edge. Abbiamo esplorato tre diversi algoritmi per ottenere questo sistema distribuito, concentrandoci in particolare sulla creazione della overlay network di broker MQTT. Le fasi dell'evoluzione di MQTT sono state testate in diversi scenari analizzandone le prestazioni, quali l’end-to-end delay, il consumo di risorse e di banda. In conclusione, questo lavoro mostra le capacità dei servizi IoT abilitati dall’edge computing e cerca una soluzione per sistemi di comunicazione efficienti per server distribuiti situati ai margini della rete.
Internet of Things prototypes and communication protocols for Edge network architectures
LONGO, EDOARDO
2021/2022
Abstract
Internet of Things (IoT) devices are the key enabler for the envisioned smart cities. In many application scenarios, 5G and Edge computing will work together to bring computational power and storage closer to IoT devices. Nevertheless, cloud-based IoT systems do not make use of edge computing which can grant real-time communication between devices, greater data privacy and location awareness support. In addition, many communication protocols still rely on centralised systems instead of the envisioned distributed architecture at the network’s edge. In this Ph.D. thesis, we study two aspects of the convergence between IoT and 5G edge computing. First, designing and implementing two IoT applications that offload computational-intensive tasks to the edge. Second, proposing a distributed publish-subscribe communication protocol for IoT edge computing nodes. Special attention is paid to the MQTT pub/sub protocol and the role of communication between MQTT brokers. In our first contribution, we propose a combined solution of two IoT prototypes for a smart campus scenario, called Smart Waste Bin and Smart Gate. To support complex decision algorithms and information fusing, the device intelligence runs on the edge of the 5G network rather than on a cloud server or locally on the devices themselves. We discuss the entire design of the system prototypes, from the analysis of requirements to the implementation details. Results indicate that moving the artificial intelligence to the edge of the network is beneficial from latency and energy consumption perspectives. The second contribution presents a new benchmarking framework for distributed MQTT brokers. The system, called BORDER, helps to evaluate the general performance of MQTT brokers beyond the envisioned application domain. Finally, in our main contribution, we propose an evolution of the MQTT protocol. Currently, it is based on a single-server topology that does not scale well considering the massive numbers of IoT devices expected in the next future. In this thesis, we aim to implement a distributed MQTT broker system to increase failure recovery, network scalability, and, thus, make MQTT even more effective in the IoT and edge domain. We explore three different algorithms to achieve such a distribution, focusing particularly on the creation of the MQTT brokers overlay network. The stages of the MQTT evolution are tested in different scenarios and performance metrics such as end-to-end delay, resource and bandwidth consumption are analysed. In conclusion, this work shows the capabilities of edge-enabled IoT devices and a solution for efficient communication systems for distributed servers located at the edge of the network.File | Dimensione | Formato | |
---|---|---|---|
main.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: thesis
Dimensione
13.43 MB
Formato
Adobe PDF
|
13.43 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/189601