OPTIMIZATION MODEL FOR COMPUTING LOAD BALANCING IN HYBRID IoT NETWORKS

Abstract

Context. The problem of optimizing computational load balancing in hybrid IoT networks combining cloud, edge, and embedded nodes under limited resources and dynamic topology conditions is addressed. The object of the study is the processes and mathematical models of load balancing in heterogeneous IoT environments.Objective. The goal of the work is to develop and experimentally verify an optimization model for computational load balancing in hybrid IoT networks, considering resource heterogeneity, dynamic changes, and energy constraints, to improve performance, scalability, and energy efficiency.Method. The study proposes an adaptive optimization model formalized as a constrained optimization problem. The model uses Lagrange multipliers, queueing theory (M/M/1), and heuristic and stochastic methods to determine optimal load distribution. The OMNeT++ simulation environment, with INET Framework and Castalia libraries, was used to test the model across 10 scenarios with varying load levels (low/mid/high), considering metrics such as processing time, node load, channel usage, and energy consumption. The model was compared to baseline methods: round-robin, random assignment, and centralized load balancing.Results. Experimental results demonstrate that the proposed model reduces processing time by 20–35%, improves average node utilization up to 90%, and decreases energy consumption by 15–20% compared to baseline methods. Simulation results were validated with multiple runs and visualized using OMNeT++ output.Conclusions. The scientific novelty of the obtained results lies in the development of an adaptive optimization model that, for the first time, comprehensively accounts for hybrid IoT network heterogeneity, energy constraints, and dynamic topology. This model provides a formal mathematical framework and practical algorithm for efficient load balancing. The practical significance of the obtained results is that the model can be directly implemented in real hybrid IoT systems, especially where energy constraints and latency requirements are critical. The developed software and simulation setup offer a ready-to-use tool for improving scalability and adaptability. Prospects for further research include extending the model to multi-channel and multi-service scenarios, incorporating fault tolerance mechanisms, and developing hybrid management strategies combining centralized and decentralized control.