命名数据网络中的一种主动拥塞控制机制研究  

Research on Active Congestion Control Mechanism in Named Data Network

作　　者：王亚东[1] 张悦[1] 陈延祥 张宇[1] WANG Yadong;ZHANG Yue;CHEN Yanxiang;ZHANG Yu((Research Institute of Communication Technology,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]北京理工大学通信技术研究所

出　　处：《载人航天》2020年第1期69-75,共7页Manned Spaceflight

基　　金：载人航天领域预先研究项目(060301,060501)

摘　　要：命名数据网络由于缓存的作用及多路径和多播的传输特点,使得传统的拥塞控制策略不再适用。从命名数据网络转发策略的角度出发,结合强化学习中的Sarsa(λ)算法,提出了一种以最小时延为目标的拥塞控制算法。该算法考虑了链路延迟和中断的影响,利用NDN中路由节点的计算和学习能力,使用Sarsa(λ)算法实现命名数据网络中网络包的智能转发。在基于ns-3的ndnSIM仿真平台下进行性能测试,并和已有的采用滑窗机制的Best route算法、Multicast算法和RF算法做比较。仿真结果表明,提出的智能转发策略能有效增加网络的数据递交率,减少丢包数量和网络平均时延,有效地减少拥塞。In named data network,the traditional congestion control strategy is no longer applicable because of the role of cache and the transmission characteristics of multipath and multicast.From the perspective of forwarding strategy in named data networks,a congestion control algorithm with minimum response time as the goal was proposed,which combined Sarsa(λ)algorithm in reinforcement learning.In the algorithm,the influence of link delay and interruption were considered,the computing and learning ability of routing node in NDN were utilized,and Sarsa(λ)algorithm was adopted to realize the intelligent forwarding of network packets in NDN.In the simulation platform of ndnSIM based on ns-3,the performance of the proposed algorithm was tested and then compared with the existing Best Route Algorithm,Multicast Algorithm and RF Algorithm which adopted the sliding window mechanism.The simulation results showed that the proposed intelligent forwarding strategy could effectively increase the data delivery rate of the network,reduce the number of packet loss and the average network delay,and reduce congestion.

关 键 词：命名数据网络 拥塞控制 强化学习 Sarsa(λ)算法 

分 类 号：TN927.3[电子电信—通信与信息系统]