基于广域无损网络的QoS参数自适应优化技术应用研究

摘要：随着生成式人工智能（AIGC）大模型计算应发展，当前广域网基于DiffServ模型的主动队列管理（AQM）技术，难以满足业务在高吞吐，低时延以及低抖动下的不同流量、多样化的业务场景需求。研究了广域网络流量的场景分类技术，并结合AQM技术，提出了一种广域服务质量（QoS）参数自适应优化方法，实现跨数据中心的无损转发，优化吞吐量和时延性能。

关键字：拥塞控制；主动队列管理；服务质量；概要统计

Abstract: With the development of generative artificial intelligence (AIGC) large model computing, the active queue management (AQM) technology based on the DiffServ model in current wide area networks can hardly meet the requirements of services for different traffic and diverse business scenarios under high throughput, low latency, and low jitter. The traffic scenario classification technology for wide area networks is studied, and combined with AQM technology, an adaptive optimization method for wide area quality of service (QoS) parameters is proposed to achieve lossless forwarding across data centers and optimize throughput and latency performance.

Keywords: weighted random early detection，queue management，quality of service，classification techniques