单源多径路由网络拥塞链路识别

针对多径路由带来的端到端测量路径不确定性以及布尔模型不能很好地解决多拥塞链路的问题，该文在识别端到端测量路径的基础上，提出一种基于扩展状态空间的网络拥塞链路识别算法。首先基于探测流时延相关性进行自适应聚类，进而得到各路径与探测流之间的映射关系。其次采用多门限的方式，将具有不同丢包程度的拥塞路径赋予不同的拥塞状态。最后将拥塞链路识别问题转化为一个约束最优化问题，并提出基于扩展状态空间的拥塞链路识别算法(ESSCLI)算法求解该问题。仿真结果表明，ESSCLI算法能够在多种不同网络场景下取得比当前算法更高的拥塞链路检测率。

Congestion Link Identification under Multipath Routing for Single-source Networks

Regarding the uncertainty introduced by load balancing when determining which end-to-end path is measured and that the classical Boolean model is not well developed for the scenario of multiple congestion links, this paper bases on the identification of end-to-end probing paths and proposes an enlarged state space based congestion link identification algorithm. Firstly, the mapping relationship between the probing flows and the measured paths is obtained after performing adaptive clustering on the probing flows with their delay correlation measures. Secondly, with multiple thresholds, it is able to assign a path with a different congestion state according to its different loss rate levels. Lastly, the issue of the congestion link identification is modeled as a constrained optimization problem, and is solved with Enlarged State Space based Congestion Link Identification (ESSCLI) algorithm. The simulation results demonstrate that ESSCLI can achieve a better detection rate of the congestion link in various network scenarios compared with existing algorithms.