基于神经网络算法的 ICN 网络传输控制研究

随着全球数字媒体的深刻变革，互联网用户关注的重点逐渐向如何快速获取信息转移，而不关注信息的存储位置。现在的 TCP／IP 网络协议架构却无法适应当今内容应用需求的迅速发展。为了适应这一互联网的转变，以信息为中心的新型网络架构信息中心网络（information-certric networking，ICN）受到了广泛关注。网络时延的动态变化反映了网络路径的负载特征，对时延的精确预测是实施网络拥塞控制、路由选择的重要依据。在 ICN 中，由于网络缓存机制导致时延的不确定性，为网络传输控制带来了挑战。通过对 ICN 网络的经典架构命名数据网络（named data networking，NDN）网络时延模型进行建模，采用了神经网络算法进行时延预测，设计了基于预测时延的转发策略机制，创新地在 NDN 网络组件转发信息表（forwarding information based，FIB）上新增接口信息 Stat，以实现转发路径的动态选择。仿真结果表明，该设计机制能够有效地提升网络传输控制性能。

Research for transport control in ICN based on neural network algorithm

With a profound change in the global digital media, current Internet users gradually show great concern for how to get information quickly, rather than the storage location of information. However,the TCP/IP network protocol architecture is unable to adapt to the rapid development of the application demand today. Thus, the new network architecture of information-centric (information-centric networking,ICN) has received the widespread attention. The dynamic changes of the network time delay reflects the load characteristics of the network path, so the accurate prediction of time delay is important to the implementation of network congestion control and routing. In ICN, the uncertainty of time delay, due to network caching mechanism, brings new challenges for the network transmission control. Therefore, in the NDN (named data networking) network which belongs to ICN delay model modeling, time delay prediction algorithm and neural network are used in this paper to design the forwarding strategy based on the predicted time delay mechanism, adding interface Stat into component forwarding information based in NDN innovatively, so as to realize the dynamic choice of forwarding path. The simulation results show that the designed mechanism can effectively improve network transmission control performance.