基于四元分组测量的网络拓扑推断算法

针对在基于层析成像技术的拓扑推断中测量节点性能时大都需要节点间的时钟同步,并且受到网络负载的影响比较大的问题,提出了一种可以测量时延抖动和丢包率的四元分组序列测量方法,并设计了根据2个网络性能参数计算节点相关性的拓扑推断算法.理论分析与仿真实验证明,该算法受到网络负载的影响较少.

A Network Topology Inference Algorithm Based on 4-tuple Packets Measurement

When measuring network performance characteristics in network topology inference,the time synchronization is needed.Meanwhile,the time network topology inference suffers the limits of network load,so there are more limits when applying network topology inference.In order to conquer the limits,a 4-tuple packets measurement method is put forward that could measure the end to end loss and jitter.An algorithm of computing the correlation between nodes is proposed which combining jitter and end to end loss.The 4-tuple packets measurement and the algorithm of computing correlation are suited for the network of lightly loaded,moderately loaded and highly loaded.The measurement method and topology inference algorithm have less limits than other algorithms.