The paper addresses the problem of optimal traffic redistribution in hybrid SDN networks that combine traditional routing technologies (OSPF) with Software-Defined Networking (SDN). It is shown that classical load balancing methods (DNS-Round Robin, ECMP, ADC, etc.) lack the required flexibility in complex telecommunication environments. A survey of recent approaches highlights the use of neural networks, reinforcement learning, multipath and energy-efficient balancing. For hybrid networks, special attention is given to both Quality of Service (QoS) requirements and hardware constraints of switches (TCAM). The problem is formalized as a Multicommodity Constrained Flow Splitting (MCFS) optimization task considering link capacities, path limitations, and memory constraints of network devices. Since the exact solution is NP-hard, an approximate metaheuristic method – Iterative Relaxation with Scaling and Rounding (IRSR) – is proposed. The algorithm combines solving a relaxed linear program, scaling of link capacities, and iterative rounding with resource reallocation. The obtained results demonstrate that IRSR effectively reduces maximum link utilization and improves bandwidth resource efficiency. The proposed approach enables automated traffic management, prevents network overloads, and supports adaptive routing policies. Future research will focus on integrating intelligent forecasting and decision-making techniques to enhance automation and performance in hybrid SDN environments.
Ustinov et al. (Wed,) studied this question.