基于探测包群的单播链路丢包率估计

Link Loss Inference in Unicast Network Based on Multi-Packet Stripe

针对单播链路丢包率估计问题,在现有方法的基础上,给出了探测包群的方法,在保证测量有效性的前提下,极大地减少了探测包的发送数量;同时提出网络层析成像的三叉树模型,增加了丢包率估计所使用的方程数目。仿真结果验证了三包组方法及三又树模型,提高了求解的精度和稳定性。

Network tomography is an edge-to-edge mechanism to infer per-link characteristics of a network domain. In this paper, we explore the use of end-to-end unicast traffic as measurement probes to infer link-level loss rates. We design experiments based on multi-packet stripe and three-leaf tree model. The purpose of these methods is to decrease the number of probes and increase the equations. Compute Simulation result show the validity of the proposed method.