A great variety of Internet of Things (IoT) applications have been developed in the last few years and are getting more and more popularity thanks to their capability of providing society and citizens with improved services and lifestyle. Many applications, ranging from autonomous driving, healthcare, smart cities up to industrial environments, increasingly exploit IoT-based services to support decision-making systems and take advantage of data-driven services. IoT applications are based on the presence of geographically distributed sensors to collect heterogeneous kinds of data about the surrounding environment. Such data are typically sent to a cloud computing data center to be processed using machine learning and Artificial Intelligence (AI) algorithms.

For some applications, the traditional cloud-based approach may not represent the most convenient choice. For example, for bandwidth-bound applications, the increasing volume of produced data is likely to make transferring and processing data at a remote cloud data center too expensive or not convenient for network constraints. Indeed, a high network utilization may be unwanted due to costs related to the cloud pricing options, even when the network load does not lead to a performance degradation. Other IoT applications have essential requirements related to latency and response time (e.g., real-time contexts) and cannot cope with a remote cloud data center’s high network delay. In these cases, an Edge Computing paradigm is likely to represent a preferable solution.

The main feature of an edge computing infrastructure is a layer of edge nodes located on the network edge, close to the sensors, to host tasks aimed at preprocessing, filtering, and aggregating the data coming from the distributed sensors. The layer of edge nodes, placed in an intermediate position between sensors and the remote cloud data center, presents a twofold advantage. First, it may reduce the data volume transferred to the cloud through pre-processing and filtering performed on the network edge. Second, the intermediate layer of edge nodes reduces latency and response times for latency-bound applications. However, the increased complexity due to the introduction of the edge nodes layer opens novel issues concerning the infrastructure design.

Date: Thursday, 23^rd February 2023

Time: 8.45 am to 12:00 pm