时间敏感网络中的可变长整形队列调整算法  

Variable-length Shaping Queue Adjustment Algorithm in Time-sensitive Networks

作　　者：蔡嫦娟 庄雷[2] 杨思锦[2] 王家兴 阳鑫宇 CAI Changjuan;ZHUANG Lei;YANG Sijin;WANG Jiaxing;YANG Xinyu(School of Cyber Science and Engineering,Zhengzhou University,Zhengzhou 450002,China;School of Computer and Artificial Intelligence,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学网络空间安全学院,郑州450002 [2]郑州大学计算机与人工智能学院,郑州450001

出　　处：《计算机科学》2024年第8期354-363,共10页Computer Science

基　　金：河南省重大科技专项(221100210900-03)。

摘　　要：针对异步整形器(ATS)采用固定长度整形队列实现流量整形存在缓存资源利用率低、可调度流平均时延高等问题,提出了一种基于改进磷虾群算法与流量预测的可变长整形队列调整算法。综合考虑流的队列分配规则、有界时延需求及有限缓存资源,定义时间敏感网络中可调度流传输约束。引入混沌映射、反向学习与精英策略并设计自适应位置更新策略以提升传统磷虾群算法的求解能力,利用改进磷虾群算法寻找整形队列可调整上限。基于卷积神经网络与长短期记忆模型(CNN-LSTM)预测流量,根据预测值计算队列长度调整步幅。仿真结果表明,与采用固定长度整形队列的方法相比,所提算法能有效提高可调度流数量,降低调度流(ST)平均时延,并提升网络缓存资源利用率。A variable length shaping queue adjustment algorithm based on an improved krill herd algorithm and traffic prediction is proposed to address the issues of low buffer resource utilization and high average delay of schedulable streams using fixed length shaping queues for traffic shaping in asynchronous traffic shaper(ATS).Considering the queue allocation rules of flows,bounded delay requirements,and limited buffer resources,transmission constraints for schedulable flows are defined in time-sensitive networks.The improved krill herd algorithm is used to find the maximum adjustable upper limit of the shaping queue,using a combination of chaos mapping,opposition-based learning,elite policy,and adaptive location update strategy to enhance the algorithm’s solving ability.The traffic is predicted based on convolutional neural network and long short-term memory model(CNN-LSTM),and the queue length is calculated according to the predicted value to adjust the step.Simulation results show that compared with the method of using fixed-length shaping queues,the proposed algorithm can effectively increase the number of sche-dulable flows,reduce the average delay of scheduled traffic(ST),and improve the utilization rate of network buffer resources.

关 键 词：时间敏感网络 异步整形器 改进磷虾群算法 流量预测 可变长队列 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]