Due to the massive amount of data generated by very pervasive IoT and mobile devices combined with a high transfer rate and real-time requirements of emerging scenarios, Cloud computing is showing some limitations. In this sense, post-cloud computing solutions (e.g., Edge and Fog computing) move part (or all) of the computation closer to the data source, making them a very hot research topic. Even if there are some tentative frameworks and standardization proposals, there are no homogeneous architectural models to integrate the various paradigms or, in many cases, they are based on proprietary solutions. Moreover, current solutions implement in part (or not at all) fine-grained resource management techniques, which are necessary to deal with energy-constrained devices. For these reasons, this thesis proposes the BarMan framework as a cooperative approach to integrate different run-time managed levels of resources from the Cloud to Edge continuum. However, dealing with the aforementioned distributed and multi-level systems means having different kinds of resource heterogeneity. In this sense, this research aims to propose a novel resource-aware and task-based programming model to overcome the current state-of-the-art limitations. Thus, since the correlation between heterogeneity and post-cloud scenarios, the programming model has also been extended for developing and integrate distributed applications. The programming model can integrate applications with the BarbequeRTRM, which monitors and manages devices resources. Furthermore, in the Fog scenario, mobile devices become part of the computing system because they can be exploited by lower-level or nearby devices to perform part of the computation. On the other hand, mobile devices increase their computational power, still being affected by their energy budget limitation. In this regard, the rest of this research work aims to enable efficient integration of mobile devices at the Fog level through the run-time management of the application's execution, device's resources allocation, and energy consumption while considering the application's performance and requirements. Finally, to meet the research community demand for real use-cases and hardware test-beds, we applied our approach to a developed video surveillance application and to a large-scale emergency scenario, evaluating them on self-built Fog test-beds and through simulation tools.
Il progressivo aumento dei dati, generati dai sempre più pervasivi dispositivi mobili e IoT unito ai requisiti real-time sempre più stringenti degli scenari emergenti, mostra il limite tecnologico dell'infrastruttura basta sul Cloud. Per superare queste limitazioni, le più recenti e interessanti soluzioni (Edge e Fog computing) spostano parte della computazione più vicino alle sorgenti di dati. Sebbene in letteratura vi siano proposti dei framework, questi molte volte non presentano modelli architetturali omogenei o sono basati su soluzioni proprietarie. Inoltre, questi ritrovati spesso implementano solo in parte una gestione fine delle risorse, necessaria per integrare dispositivi con energia limitata. Dunque questa tesi presenta BarMan come un framework per integrare senza soluzione di continuità differenti livelli di risorse dalle piattaforme Cloud ai dispositivi edge. Tuttavia un sistema così distribuito e multilivello richiede di gestire risorse eterogenee. A tal proposito questo lavoro presenta un innovativo modello di programmazione, basato su applicazioni modulari e integrato con il sistema di gestione delle risorse BarbequeRTRM, e di distribuzione trasparente dei moduli. Negli scenari di Fog computing, i dispositivi mobili possono essere utilizzati da altri dispositivi vicini per svolgere parte della computazione. Nonostante ciò, sebbene negli anni essi abbiano aumentato la loro potenza computazionale, continuano ad avere vincoli energetici. Pertanto la restante parte della ricerca si è concentrata sull'integrazione efficiente di questi dispositivi nell'infrastruttura Fog, attraverso la gestione a run-time dell'esecuzione delle applicazioni, dell'allocazione delle risorse e della potenza consumata, tenendo conto delle prestazioni e dei requisiti della singola applicazione. Infine, data la necessità nella comunità scientifica di avere scenari di applicazione reali, abbiamo adattato il nostro approccio ad uno scenario di videosorveglianza e ad un sistema di gestione delle emergenze, valutandoli su dispositivi reali e attraverso software di simulazione.
Post-Cloud Computing: Addressing Resource Management in the Resource Continuum
Zanella, Michele
2021/2022
Abstract
Due to the massive amount of data generated by very pervasive IoT and mobile devices combined with a high transfer rate and real-time requirements of emerging scenarios, Cloud computing is showing some limitations. In this sense, post-cloud computing solutions (e.g., Edge and Fog computing) move part (or all) of the computation closer to the data source, making them a very hot research topic. Even if there are some tentative frameworks and standardization proposals, there are no homogeneous architectural models to integrate the various paradigms or, in many cases, they are based on proprietary solutions. Moreover, current solutions implement in part (or not at all) fine-grained resource management techniques, which are necessary to deal with energy-constrained devices. For these reasons, this thesis proposes the BarMan framework as a cooperative approach to integrate different run-time managed levels of resources from the Cloud to Edge continuum. However, dealing with the aforementioned distributed and multi-level systems means having different kinds of resource heterogeneity. In this sense, this research aims to propose a novel resource-aware and task-based programming model to overcome the current state-of-the-art limitations. Thus, since the correlation between heterogeneity and post-cloud scenarios, the programming model has also been extended for developing and integrate distributed applications. The programming model can integrate applications with the BarbequeRTRM, which monitors and manages devices resources. Furthermore, in the Fog scenario, mobile devices become part of the computing system because they can be exploited by lower-level or nearby devices to perform part of the computation. On the other hand, mobile devices increase their computational power, still being affected by their energy budget limitation. In this regard, the rest of this research work aims to enable efficient integration of mobile devices at the Fog level through the run-time management of the application's execution, device's resources allocation, and energy consumption while considering the application's performance and requirements. Finally, to meet the research community demand for real use-cases and hardware test-beds, we applied our approach to a developed video surveillance application and to a large-scale emergency scenario, evaluating them on self-built Fog test-beds and through simulation tools.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_Thesis_Zanella.pdf
non accessibile
Descrizione: Thesis
Dimensione
12.2 MB
Formato
Adobe PDF
|
12.2 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/10589/188757