Load-Tolerant Many-Objective Optimization for Resource Allocation in SDN-based IoT

Abstract

The increasing workload in Internet of Things (IoT) environments leadsto network congestion and resource imbalance, significantly degradingthe Quality of Service (QoS). Software-Defined Networking (SDN)provides a flexible control paradigm for improving QoS and loadbalancing. However, most existing approaches are limited to multi-objective formulations and do not explicitly address overloadconditions. Therefore, there is a need for many-objective QoSoptimization frameworks that can jointly consider multiple conflictingobjectives while maintaining load tolerance in SDN-based IoT systems.In this paper, a load-tolerant many-objective resource allocationframework is proposed to simultaneously optimize user and serviceprovider QoS requirements. Specifically, cost and response time areminimized for users, while energy consumption is minimized andresource utilization is maximized for service providers. A Pareto-basedmany-objective evolutionary optimization process is employed togenerate diverse non-dominated solutions, which are evaluated underoverload conditions. To address infeasible allocations under overload, aload-tolerance mechanism is introduced to identify admissible solutionsand improve system robustness. This mechanism enables sustained taskallocation even when Pareto-optimal solutions violate load threshold,thereby increasing the task acceptance rate under overload conditions.Simulation results demonstrate that the proposed Many-objective GreyWolf Optimization (MaGWO)-based framework improves resourceallocation cost by 13.85%, response time by 17.2%, energy consumptionby 15.8%, and resource utilization by 10.25%. Furthermore, theproposed load-tolerant strategy increases the task acceptance rate by8.4% compared to NSGA-III and Many-objective Particle SwarmOptimization (MaPSO).