Evaluation of Large Scale RoI Mining Applications in Edge Computing Environments

Researchers and leading IT companies are increasingly proposing hybrid cloud/edge solutions, which allow to move part of the workload from the cloud to the edge nodes, by reducing the network traffic and energy consumption, but also getting low latency responses near to real time. This paper proposes a novel hybrid cloud/edge architecture for efficiently extracting Regions-of-Interest (RoI) in a large scale urban computing environment, where a huge amount of geotagged data are generated and collected through users's mobile devices. The proposal is organized in two parts: (i) a modeling part that defines the hybrid cloud/edge architecture capable of managing a large number of devices; (ii) a simulation part in which different design choices are evaluated to improve the performance of RoI mining algorithms in terms of processing time, network delay, task failure and computing resource utilization. Several experiments have been carried out to evaluate the performance of the proposed architecture starting from different configurations and orchestration policies. The achieved results showed that the proposed hybrid cloud/edge architecture, with the use of two novel orchestration policies (network- and utilization-based), permits to improve the exploitation of resources, also granting low network latency and task failure rate in comparison with other standard scenarios (only-edge or only-cloud).