A Shared Bottleneck Aware Multi - Path Selection Scheme for NDN

Yong Wang, Wang Yang, Jieyu Zhou

Abstract: Different from the traditional TCP/IP network, the Named Data Networking (NDN) naturally supports consumers to obtain content from multi-source through multi-path. In NDN, the number of available sub-paths of consumers is significantly increased, which means that NDN has the potential to provide better multi-path transmission. However, excessive sub-path use can create shared bottlenecks, leading to increased queuing delays. Also, shared bottleneck detection is not an easy task due to the path lag problem. Therefore, we propose SBAMPS, a Shared Bottleneck Aware Multi-Path Selection scheme designed for NDN. SBAMPS contains CDDTW-SBD, a Shared Bottleneck Detection algorithm based on Congestion Degree (CD) and Dynamic Time Warping (DTW). CDDTW-SBD defines the max queuing ratio of intermediate nodes on the sub-path as its CD, and then calculates the CD similarity of every two sub-paths based on DTW to determine whether they share a bottleneck. Afterward, SBAMPS selects a better sub-path among every two shared bottleneck sub-paths based on their performance evaluation. Simulation results show that CDDTW-SBD can mitigate the interference of path lag and improve detection accuracy by an average of 26 % as compared to the benchmark. SBAMPS can reduce the latency by up to 47.13% and achieve comparable throughput as compared to benchmarks not considering shared bottleneck.

External IDs:dblp:conf/hpcc/WangYZ23