Vitale, Vincenzo Norman (2022) TOWARDS COST-PERFORMANCE AWARENESS IN FOG-BASED DATA MANAGEMENT ARCHITECTURES. [Tesi di dottorato]

Preview

Text VITALE_VincenzoNorman_34.pdf Download (38MB) | Preview

Collection description

The evolution and diffusion of the latest Information and Communication Technologies (ICT) allowed for the constitution of huge intelligent environments to improve and facilitate users’ lives by providing innovative services and information. Underlying these environments is the Internet of Things (IoT), which has been identified as the key driver for implementing intelligence in any environment, such as factories, cities, or homes. Such a rapid technological evolution combined with easy access to an internet connection had various effects. First, it dramatically increased the amount of data collected by IoT sensor networks in the form of time series. Then, it made it easier for users to access data and services. This has put a strain on the DBMSs and the existing architectures designed to manage IoT data flows, resulting in increased latencies and network traffic. New databases for temporal data management have been designed and combined with consolidated Cloud-centric architectures to face the problems mentioned above. Nevertheless, as Cloud computing demonstrated its limits in the presence of time-sensitive tasks, new computational paradigms like Fog computing have been introduced to overcome the limits deriving from Cloud-centric data management. Introducing new solutions based on these technologies has brought numerous advantages and challenges. For example, classic distributed data management approaches have proven ineffective in some cases, leading to cost/performance inefficiencies. Likewise, abuse or misuse of the Fog paradigm can lead to resource waste or even performance deterioration. The objective of the thesis is to provide a hybrid Cloud-Fog architecture, which allows reducing the costs of data management compared to Cloud-based architectures without excessively reducing performance. To this end, a Fog level is introduced, which assumes a more active role in data management, to use the intrinsic knowledge of the area served and the related analytical workloads. The effectiveness of the architecture is shown using real data and typical analytical configurations with state-of-the-art databases. However, we designed and developed an extensible simulation tool due to the lack of appropriate tools for evaluating similar architectures. This tool aims to provide a solution architect with the basic functionalities both to evaluate the introduction of a Fog layer and evaluate the efficiency of the different strategies. Finally, we propose a new evaluation metric that provides an overview of the cost-performance efficiency for the considered architecture.

Downloads

Actions (login required)

View Item