QoSCAPE: QoS-Centric Adaptive Path Engineering with Blockchain-Enabled Reinforcement Learning

Abstract

In the evolving landscape of effective network management, path engineering, also known as traffic engineering, optimization and selection remain a critical challenge. This paper presents an innovative framework that integrates Software Defined Networking (SDN), Blockchain technology (BC), and Reinforcement Learning (RL) to enhance the efficiency and security of network path optimization. SDN's centralized control enables dynamic network traffic management, while BC ensures secure, transparent, and immutable logging of network transactions, fostering trust and accountability. Reinforcement learning, capable of learning and adapting from environmental interactions, is employed to dynamically optimize routing decisions. Our proposed framework, QoSCAPE: QoS-Centric Adaptive Path Engineering with Blockchain-Enabled Multi-Agent Reinforcement Learning, leverages the programmability of SDN to collect real-time network state information and the decentralization of BC to secure this data. The numerical results underscore the superior efficiency of QoSCAPE, which consistently achieves near-total request success rates within milliseconds, significantly outperforming traditional HRA and DRA methods in rapidly fulfilling network service demands in terms of Path Setup Time (PST) and Requests Serviced (RS) metrics, yet notably surpasses all other approaches by minimizing Network Message Overhead (NMO). Its optimization in reducing message volume ensures efficient resource usage and preserves network scalability, distinguishing it as a superior choice for multi-ISP routing frameworks. © 2024 Elsevier B.V., All rights reserved.