METHOD OF CONTROLLING THE SD-WAN COMPUTER NETWORK USING MACHINE LEARNING METHODS BASED ON A MATHEMATICAL MODEL IN THE STATE SPACE

DOI: 10.31673/2412-4338.2026.019020

Abstract

The development of computer network construction technologies has led to the creation of software-defined wide area networks (SD-WAN). This direction is one of the key trends in the development of effective network infrastructure. In this case, software centralized management of the entire distributed computer network is provided using a specialized controller that separates the control system from the data transmission system.

The purpose of this work is to improve the efficiency of functioning of distributed computer networks SD-WAN using machine learning methods based on a mathematical model in the state space. The work solves the current scientific problem of developing a method for controlling a computer network SD-WAN based on machine learning in the state space. A mathematical model of the SD-WAN network in the state space has been developed, which represents a set of state vectors, control and a quality of service function. It takes into account the dynamics of channel loading, delay, packet loss and the state of node buffers. This model can be used in the form of a linearized version for analytical calculations, and in the form of a full nonlinear form for simulation. The conditions for stability and controllability of a computer network are determined. For the linear model, an analytical solution to the optimal SD-WAN control problem is obtained, which is based on the Ricchati equation. A network control algorithm is developed based on deep reinforcement learning. To study the effectiveness of the obtained results, a simulation of a distributed computer network using different SD-WAN management methods was conducted. The simulation results confirmed the superiority of the proposed state-space machine learning-based SD-WAN computer network management method. They showed a 65% reduction in latency, an 85% reduction in packet loss, and a 61% improvement in the quality functional compared to the basic ECMP method.

Keywords: distributed computer network, SD-WAN, machine learning methods, mathematical model, state space method, management, quality of service, communication channel.

References

[1] Gnidenko M.P., Vyshnivskyi V.V., Zinchenko O.V., Ishcheryakov S.M. A new dimension of implementing key functions and capabilities of the latest HPE Aruba technologies / Monograph / – Kyiv: DUIKT, FOP Gulyaeva V.M., 2025. – 289 p.

[2] Meyer D. Software-Defined Networking: A Comprehensive Survey / D. Meyer, J. Clarke, I. Moraes. – Proceedings of the IEEE. – 2015. – Vol. 103, No. 1. – P. 14–76.

[3] Xiao Y. Reinforcement Learning-Based QoS/QoE-Aware Service Function Chaining in Software-Defined and Virtualized Multimedia Networks / Y. Xiao, Q. Zhang // IEEE Transactions on Multimedia. – 2017. – Vol. 19, No. 7. – P. 1555–1564.

[4] Liu S. Deep Reinforcement Learning for Dynamic Multichannel Access in Wireless Networks / S. Liu, W. Wang // IEEE Transactions on Cognitive Communications and Networking. – 2018. – Vol. 4, No. 2. – P. 257–265.

[5] K. Pentikousis et al., “Software-Defined Networking (SDN): Layers and Architecture Terminology,” RFC 7426, Jan. 2015.

[6] X. Wu, K. Lu, and G. Zhu, “A Survey on Software-Defined Wide Area Networks,” Journal of Communications, vol. 13, no. 5, pp. 253-258, 2018, doi: 10.12720/jcm.13.5.253-258.

[7] S. Jain et al., “B4: Experience with a Globally-Deployed Software Defined WAN,” ACM SIGCOMM Computer Communication Review, vol. 43, no. 4, pp. 3-14, 2013.

[8] C. Hong et al., “Achieving High Utilization with Software-Driven WAN,” ACM SIGCOMM Computer Communication Review, vol. 43, no. 4, pp. 15-26, 2013.

[9] A. Gupta et al., “SDX: A Software Defined Internet Exchange,” ACM SIGCOMM Computer Communication Review, vol. 44, no. 4, pp. 551-562, 2015. DOI:10.1145/2740070.2626300

[10] M. L. Puterman, Markov Decision Processes: Discrete Stochastic Dynamic Programming. New York, NY, USA: Wiley, 1994.

[11] A. Rantzer, “On the Kalman-Yakubovich-Popov lemma,” Systems & Control Letters, vol. 28, no. 1, pp. 7-10, 1996.

[12] J. B. Rawlings, D. Q. Mayne, and M. Diehl, Model Predictive Control: Theory, Computation, and Design, 2nd ed. Madison, WI, USA: Nob Hill Publishing, 2017.

[13] Mestres A. Knowledge-Defined Networking / A. Mestres, A. Rodriguez-Natal, J. Carner et al. // ACM SIGCOMM Computer Communication Review. – 2017. – Vol. 47, No. 3. – P. 2–10. 

[14] Mnih V. Human-level control through deep reinforcement learning / V. Mnih, K. Kavukcuoglu, D. Silver et al. // Nature. – 2015. – Vol. 518. – P. 529–533.

[15] Stampa G. A Deep-Reinforcement Learning Approach for Software-Defined Networking Routing Optimization / G. Stampa, M. Arias, D. Sanchez-Charles et al. – arXiv:1709.07080. – 2017.

[16] Schulman J. Proximal Policy Optimization Algorithms / J. Schulman, F. Wolski, P. Dhariwal, A. Radford, O. Klimov. – arXiv:1707.06347. – 2017.