Scheduling the execution of tasks at the edge

Abstract

The Internet of Things provides a huge infrastructure where numerous devices produce, collect and process data. These data are the basis for offering analytics to support novel applications. The processing of huge volumes of data is a demanding process, thus, the power of Cloud is already utilized. However, latency, privacy and the drawbacks of this centralized approach became the motivation for the emerge of edge computing. In edge computing, data could be processed at the edge of the network; at the IoT nodes to deliver immediate results. Due to the limited resources of IoT nodes, it is not possible to have a high number of demanding tasks locally executed to support applications. In this paper, we propose a scheme for selecting the most significant tasks to be executed at the edge while the remaining are transferred into the Cloud. Our distributed scheme focuses on mobile IoT nodes and provides a decision making mechanism and an optimization module for delivering the tasks that will be executed locally. We take into consideration multiple characteristics of tasks and optimize the final decision. With our mechanism, IoT nodes can be adapted to, possibly, unknown environments evolving their decision making. We evaluate the proposed scheme through a high number of simulations and give numerical results.