【摘要】 基于网络时延推测的2个假设、时延推测模型和路径时延数据采集方法,提出了一种基于链路重构—解构的端到端网络链路时延推测方法,应用伪似然估计将原整体问题分解为若干独立子问题分别求解,利用链路重构—解构确定可求解的推测单元,控制平均采样精度和减少推测单元链路数,从而显著降低计算复杂度。通过基于模型的计算和基于NS2的仿真实验研究,验证了推测方法的准确性和有效性。更多还原

【Abstract】 Based on two assumptions, inference model and end-to-end delay data acquisition, an approach to end-to-end network internal link delay inference based on link reconstruction-deconstruction(LRD) was proposed. Pseudo likelihood estimation(PLE) was adopted and the inference problem was divided into independent sub-problems. Inference units with definite solution are determined by LRD. By means of controlling average sampling precision and decreasing inference unit links, the computation complexity of link delay inference was significantly lowered. Experimental study was performed based on model computation and NS2 simulation platform. Theoretical analysis and experimental results show that the approach is accurate and effective.更多还原