Tprobe:基于减法模型的端到端瓶颈带宽测量方法

总结了近几年瓶颈带宽测量的技术发展历程,提出了把测量方法从考察单一组数据包对的行为,推进为研究数据包对与数据包对之间关系的新思路;建立了一个新的Two-pairs瓶颈带宽测量分析模型,通过两组数据包对之间的相减关系消除链路层协议的误差影响;基于该模型提出了一种新的测量工具Tprobe,通过采用数据重用方法,进一步提高测量数据的利用率;OPNET仿真环境下的大量实验和教育网中实际测量试验表明,Tprobe具有较好的测量准确度和健壮性,并能够有效的减少测量时间和带宽消耗.

Tprobe:a Subtraction-Based Approach for End-to-End Bottleneck Bandwidth Estimation

Bottleneck bandwidth measurement on an end-to-end path is of great importance for various network applications. Different from the existing tools that focus on the dispersion experienced by single packet-pair, a new estimation method is proposed to examine the relations among multiple packet-pairs. According to this method, a novel Two-pairs analytical model for bottleneck bandwidth estimation is presented, which can removed the effect of lower layer headers via the subtraction relation between two "differentiated packet-pairs"(DPPs). Based on the Two-pairs model, a new active probing tool to measure the bottleneck bandwidth, Tprobe, is put forward. By reusing the dispersion information of the DPPs, Tprobe can further improve the efficiency of probing. The simulations in OPNET and experiments in the China Education and Research Networks show that, Tprobe can provide more accurate and robust estimation, while using much fewer probing packets and shorter measurement latency.