|IEZZI DOMENICO||Cycle: XXXIV |
Section: Computer Science and Engineering
Tutor: AMIGONI FRANCESCO
Advisor: FORNACIARI WILLIAM Major Research topic
:Analysis and Evolution of Edge Computing SolutionsAbstract:
The proliferation of the Internet of Things, introduced for supply chain management field and expanded in many fields such as healthcare, home, environment, transports, is showing the limitation of the cloud computing paradigm. Computation is entrusted to a small number of remote data centers, pushing the bandwidth requirements to the limit due to the increasing amount of devices connected on the network. In this context, Edge Computing is emerging as a new paradigm, which calls for moving the computation of data at the proximity of data sources. i.e., the “edge” of the network. This in order to address the concerns of response time requirements, battery life constraints of mobile devices, bandwidth costs, data safety and privacy. In the Edge Computing paradigm we can perform computing, data storage, caching by distributing requests from cloud to users and improving response time as well as reducing energy consumption.
This novel paradigm brings new opportunities and challenges to be faced. Since the number of devices to the edge of the network is rapidly increasing, edge computing nodes should be able to scale efficiently and cooperatively. At this regard, due to their distributed nature, it is necessary to define the way in which nodes can communicate each other, and how workload or failovers can be managed.
To address resource and failover management on the edge nodes we can consider the introduction of a run-time resource manager, that can perform distributed resource allocation, load balancing and provide offloading strategies that take into account the power/performance trade-off, reliability and real-time requirements. In this regard, we are going to use the BarbequeRTRM, an open source resource manager developed by the Politecnico di Milano in several EU projects (2PARMA, CONTREX, HARPA, MANGO, RECIPE). Thanks to its modular nature, it can be adapted to different architectures and application scenarios, making it an evaluable choice for further developments targeting Edge Computing systems. Moreover, its support for distributed resources can be extended to leverage a fully interconnected network of devices and edge nodes, enabling resource managing optimizations in a distributed fashion.
Also Edge Computing solutions are being adopted in the Industrial IoT field, to manage sensors and devices typically deployed in the shop floor. The most important one is the open source framework EdgeX Foundry, consisting of a set of loosely-coupled microservices which allows the control of “southbound” devices, sensors and actuators as well as synchronize with “northbound” applications and cloud services.