Near real-time scheduling in cloud-edge platforms

As Cloud-Edge architectures are becoming more and more popular, due to their improvement on the battery life of the IoT devices and the high availability of data from the Cloud, this approach also creates new problems. As data gathered from the Edge has to be transferred to the Cloud in order to be processed, the result will be a decreased responsiveness of the system. Also, devices might have to process data by themselves, as the Cloud could be unreachable at random moments in time, resulting in a reduction in battery life. Therefore, we propose an architecture that solves these problems, by introducing an intermediate layer, called Fog, which uses a task scheduling algorithm to send data received from Edge to another device that has enough resources and the required hardware and software to complete the task. In addition, the architecture is based on microservices, hence improving scalability and flexibility. In the performance analysis, we used different values to find the best node that should receive the data for processing. In addition, we compared the microservice based architecture with a monolithic one in order to see how the throughput and responsiveness of the system are affected.