A Hybrid Meta-Heuristic Algorithm for Application Module Placement in IoT-Fog-Cloud Computing Environment

Abstract

Abstract In recent years, fog computing has gained significant popularity for its reduced latency (delay), low power consumption, mobility, security and privacy, network bandwidth, and real-time responses. It provides cloud-like services to Internet of Things (IoT) applications at the edge of the network with minimal delay and real-time responses. Fog computing resources are finite, computationally constrained, and powered by battery cells, which require optimal power management. To facilitate the execution of IoT services on fog computing resources, applications are broken down into a group of data-dependent application modules. The application modules communicate and transfer data from one module to another in order to achieve a common goal. With the limitations on computing resource capacity and the rise in demand for these resources for application module processing, there is a need for a robust application module placement strategy. Inefficient application module placement can result in a tremendous hike in latency, a higher completion time, a fast drain on battery cells, and other placement problems. This paper focuses on minimising the average delay, completion time (Makespan time), and energy usage of the fog system while placing the data-dependent modules of the IoT application on resources in the fog layer. To achieve the said objectives, a hybrid meta-heuristic algorithm based on the Red Deer Algorithm (RDA) and the Harris Hawks Optimisation Algorithm (HHO) is proposed. The optimisation algorithms independently search for a placement solution in the search space and update the best solution based on some probability function. The proposed hybrid algorithm was implemented using the iFogSim simulator and evaluated based on average completion time, average latency, and average energy consumption. The simulation results show the effectiveness of the proposed hybrid heta-heuristic algorithm over the traditional RDA and HHO algorithms.