Next-generation core networks are expected to achieve service-oriented traffic management for diversified Quality-of-Service (QoS) provisioning based on software-defined networking (SDN) and network function virtualization (NFV). In this article, a learning-based transmission protocol customized for Video-on-Demand (VoD) streaming services is proposed for a Cybertwin-enabled next-generation core network, which provides caching-based congestion control and throughput enhancement functionalities at the edge of the core network based on traffic prediction. The per-slot traffic load of a VoD streaming service at an ingress edge node is predicted based on the autoregressive integrated moving average (ARIMA) model. To balance the tradeoff between network congestion and throughput enhancement, a multiarmed bandit (MAB) problem is formulated to maximize the expected overall network performance in a long run, by capturing the relationship between transmission control actions and QoS provisioning. A comprehensive transmission protocol operation framework is also presented with in-network congestion control and throughput enhancement modules. Simulation results are presented to validate the efficacy of the proposed protocol in terms of packet delay, goodput ratio, throughput, and resource utilization.

中文翻译：

---

基于学习的传输协议定制，用于支持 Cyber​​twin 的下一代核心网络中的 VoD 流媒体


下一代核心网有望基于软件定义网络（SDN）和网络功能虚拟化（NFV）实现面向服务的流量管理，实现多样化的服务质量（QoS）提供。在本文中，为支持 Cyber​​twin 的下一代核心网络提出了一种为视频点播 (VoD) 流媒体服务定制的基于学习的传输协议，该协议在边缘提供基于缓存的拥塞控制和吞吐量增强功能。基于流量预测的核心网络。基于自回归综合移动平均 (ARIMA) 模型来预测入口边缘节点处 VoD 流媒体服务的每时隙流量负载。为了平衡网络拥塞和吞吐量增强之间的权衡，提出了多臂老虎机 (MAB) 问题，通过捕获传输控制操作和 QoS 供应之间的关系，最大限度地提高长期预期的整体网络性能。还提出了一个全面的传输协议操作框架，包括网络内拥塞控制和吞吐量增强模块。仿真结果验证了所提出的协议在数据包延迟、吞吐量、吞吐量和资源利用率方面的有效性。