论文部分内容阅读
因特网发展得越来越庞大,越来越复杂,各条链路的运行参数对于评估网络性能十分重要,但有时无法直接测量。网络断层成像(network tomography)技术研究基于端-端测得的数据推算链路的性能参数。本文提出了利用依赖树模型和上向——下向算法,推测链路的时延分布。仿真实验证明了这种方法的正确性,能够根据端-端测量结果精确地推算出链路的时延特性。
As the Internet grows larger and more complex, the operational parameters of each link are important to assess network performance, but sometimes can not be measured directly. Network tomography technology studies the performance parameters of a link based on end-end measured data. This paper proposes the use of dependent tree model and up-down algorithm to infer the link delay distribution. Simulation results show the correctness of this method, which can accurately estimate the link delay characteristics based on the end-to-end measurement results.