اميررضا صالحي

Software-Defined Networking (SDN) is a modern architecture for network management that separates the control plane from the data plane an‎d manages the network through a centralized controller. This structure enables dynamic routing, optimal resource management, intelligent implementation, an‎d enhanced security SDNs are utilized in cloud networks, IoT, an‎d data centers to optimize performance an‎d improve flexibility. Despite their many advantages, SDNs also face challenges such as scalability, processing delays, network resource management, dynamic allocation, an‎d ban‎dwidth prioritization to guarantee Quality of Service (QoS). QoS factors include delay, throughput, an‎d packet loss rate. In previous studies, machine learning has been employed to improve throughput, reduce delays, an‎d minimize packet loss. In large-scale networks, higher throughput an‎d lower delays can be achieved; therefore, deep reinforcement learning algorithms have been applied, yielding promising results. The proposed framework is eva‎luated based on criteria such as throughput, delay, packet loss rate, request acceptance ratio, an‎d peak signal-to-noise ratio, an‎d compared with existing solutions. Results show that the use of Deep Q-Learning not only performs better in small an‎d medium-sized networks but also brings significant improvements to QoS an‎d user experience in large-scale, complex networks — especially under high-traffic conditions. In this study, a 4 percentage improvement was achieved compared to the benchmarked method. Keywords: Software Defined Networking (SDN), Quality of Service (QoS), Quality of User Experience (QoE), Deep Reinforcement Learning (DRL)

شبكه مبتني بر نرم افزار , كيفيت خدمات , كيفيت تجربه ي كاربري , يادگيري تقويتي عميق

SoftWare Defined Network , Quality of service , Quality of experience , Deep Reinforcement Learning

Amirreza Salehi

Dr. Mozayeni