一种链路丢包门限动态变化的网络拓扑推测算法

目前端到端逻辑拓扑推测方法主要有极大似然方法和分群方法。极大似然方法的计算量会随网络规模的增加而急剧增长，从而影响在实际网络中的应用。采用计算量较小的分群推测方法，针对GLT算法中采用固定丢包率判决门限，所导致的较大推测误差，提出了改进的任意拓扑推测算法GLT。该算法利用每次迭代过程中得到的链路丢包率的估计值对，进行动态调整。仿真结果表明，GLT算法将ξ与链路丢包率估计值相结合，有效地防止了采用GLT算法导致的拓扑推测准确率的严重恶化，提高了算法性能。

A topology inference algorithm based on dynamically adapted link loss-ratio

MLE and grouping methods recently have been proposed as means to infer network logical topology, but the time spent on MLE increased sharply with the size of the networks. Aiming at the disadvantages brought by fixed in GLT algorithm, this paper proposes an improved algorithm IGLT based on the grouping method with less computation, which dynamically adapts according to the estimation of link loss-ratio. Compared with GLT algorithms, the simulation results prove that IGLT combining the estimation of link loss-ratio shows greater performance.