Edge Computing Systems: Modeling and Resource Optimization for Augmented Reality and Soft Real-time Applications

This paper presents a solution to support IoT devices by employing edge computing resources. These devices usually run tasks with high resource demands and real-time constraints such as augmented reality applications. The solution incorporates an optimization algorithm to improve the utilization of edge computing resources globally. The algorithm allocates the IoT tasks so that the number of edge computing servers is reduced, where it can handle the unpredictable and dynamic nature of edge computing systems due to its low reallocation overhead. The paper also introduces a formal representation of the considered problem and an abstraction model for edge computing systems that is compatible with the well-known EdgeCloudSim simulator. Moreover, it provides theoretical analysis for the lower bound of resource demands. Such modeling and analysis enable researchers to integrate edge computing algorithms in the simulator easily and validate the ones related to resource optimization. To evaluate the proposed solution, data sets were generated based on real execution traces of several augmented reality tasks. The evaluation shows that the proposed algorithm outperforms the state-of-the-art related techniques.