Load Balancing Based on Many-objective Particle Swarm Optimization Algorithm with Support Vector Regression in Fog Computing

With the development in computing technologies, fog computing is developing as public and robust computing, which complements cloud computing to provide services, computation, and storage on the edge network. The future growth and support of 5G access networks additional advance the viability and implementation of fog networks and widen the scope of devices that can participate and serve in IoT communication. However, scaling fog computing, the number of end-users increases. Hence, the workload between fog nodes needs to be distributed efficiently. Otherwise, some of the nodes will be overloaded, and others will be under-loaded. Consequently, one of the critical factors for managing resources in fog computing efficiently and avoiding overloaded or under-loaded is load balancing. Therefore, load balancing between these resources is a challenge in fog computing. There are different techniques to balance the load, such as optimization algorithms or machine learning. This paper proposes a load balancing model in fog computing based on a many-objective particle swarm optimization (PSO) algorithm with support vector regression (SVR). The proposed load balancing model considered four metrics to optimize them while distributing the load: response time, energy consumption, resource utilization, and throughput. Besides, It combines SVR with PSO to improve PSO performance. The proposed model has been simulated and tested to evaluate the performance from different aspects. The experiments show that the proposed model efficiently balances the load with optimizing the four metrics. In addition, it improves the performance of PSO, which is used to balance the load.