Abstract:

As a new network architecture, named data networking (NDN) takes content-centric model and conne-ctionless transmission mode. NDN is requester-driven. If the required content is too large, the data provider would slice the content and deliver each piece of content into a data packet. In this condition, the data requester keeps sending the same amount of interest packets to get the full content, which causes high delay, or the data provider returns all data packets at once, which easily results in packet loss. Thus, NDN is not applicable to large data volume and delay-sensitive applications such as AR/VR. A bandwidth-delay aware congestion control mechanism on the transmission mode of one-interest-multiple-data is firstly presented to be applied in the data provider. It calculates the congestion window and data transmission rate based on the bottleneck bandwidth on the routing path and the delay collected by the received interest packet, and sends data packets reasonably. Then, the data requester keeps sending interest packets based on the received data packets. The large data volume will be got by the data requester with a short delay. A cache tag method is then proposed. The cached content can be transmitted in order by marking the intermediate node’s cache information along the routing path, which can make full use of the cache and reduce the transmission time. Finally, the experimental results based on ndnSIM show that, compared with other congestion control mechanisms, the proposed mechanism achieves better performance, in terms of congestion window, queue length and data transmission rate. It can transmit large content with short delay. Moreover, when there are packet loss and the intermediate node cache content in the network, the proposed mechanism achieves larger performance advantage.

Key words: named data networking (NDN), congestion control, content cache, transmission mode, bottleneck bandwidth

摘要：

作为一种新型网络架构,命名数据网络（NDN）采用内容中心和无连接传输模式。NDN是内容请求者驱动,如果其所需内容较大,内容提供者将内容分片,并装入多个数据包返回。此时或者内容请求者持续发送等量兴趣包才能获取完整内容,传输延时过高;或者内容提供者一次性将数据包全部返回,易造成丢包。因此,NDN不适用于AR/VR等大数据量且延迟敏感型应用。首先基于One-Interest-Multiple-Data传输模式,提出应用于内容提供者的瓶颈链路带宽时延感知的拥塞控制机制,内容提供者通过兴趣包收集的路径瓶颈带宽和时延信息计算拥塞窗口和数据发送速率,从而合理地发送数据包,内容请求者再据此继续发送兴趣包,最终以较短延时获取大数据量内容。并进一步提出缓存标记方法,通过标记路径中中间节点的缓存信息,使得各节点缓存内容可以有序传输,充分地利用了节点缓存,减少传输所需时间。最后,基于ndnSIM的实验结果表明,与传统和改进的拥塞控制机制相比,提出的机制在拥塞窗口、队列长度和数据传输速率等方面取得更好的性能,能够以更短的时间完成大数据量内容传输。而且在网络中存在丢包和中间节点存在缓存内容时,提出的机制均达到了更好的性能优势。

关键词: 命名数据网络（NDN）, 拥塞控制, 内容缓存, 传输模式, 瓶颈带宽