异构网络多路径并行传输吞吐量建模

摘要

Based on analyzing the characteristics of concurrent multipath transfer (CMT), this paper proposes a throughput modeling method for CMT in heterogeneous networks. In order to estimate the throughput of CMT, it uses the state transition probability matrix to analyze the growth mechanisms of the congestion window in the slow-start and congestion avoidance stage of each path. It also puts forward an optimization algorithm, which reduces the complexity of the model by merging the states with the same congestion window size and adding the corresponding state transition probabilities. Experimental results show that the proposed method can accurately estimate the throughput of CMT even if the paths have different features, and that it can work well under the heterogeneous network environment.%在分析多路径并行传输特点的基础上,提出了一种异构网络多路径并行传输吞吐量的建模方法.通过引入状态转换概率矩阵对各条路径慢启动阶段和拥塞避免阶段的拥塞窗口增长机制进行分析,从而估算多路径并行传输的吞吐量；然后合并拥塞窗口值相同的状态,将对应的状态转换概率相加,从而降低了模型的计算复杂度.仿真和实验表明,即使在路径特性不同时,该方法也能准确估算多路径并行传输吞吐量的变化,在异构网络环境下具有较强的适用性.