网络编码已被验证是改善无线网络吞吐量、可靠性等性能的有效手段,但其对业务流时延性能影响的研究还不充分。针对得到广泛应用的随机线性网络编码,提出了一种基于马尔可夫状态转移的方法对单跳场景下单播业务流时延的期望值以及概率分布进行了分析,同时给出了相应的递推解析式,然后数值分析了有限域大小、信道丢失率、编码批次大小等编码参数对时延的影响。结果表明,随着有限域q的增大,时延的方差不断变小;时延的期望值与信道丢失率、编码批次大小几乎都成线性关系。
Network coding has been validated as an effective means to improve network throughput and reliability, but the study of its delay performance is not enough. Against the widely used random linear network coding, the proposed method, based on the Markov state transfer, analyzed the delay expectation and probability distribution of the unicast flow in a single hop scene, and presented the exact recursions. Then according to numerical simulation, the effect of the encoding parameters were analyzed such as the finite field size, the channel loss rate and the encoding batch size on the delay. The results show that, as the finite field q increases, the delay variance continues to become smaller; the average delay is almost linear with the channel loss rate and the encoding batch size individually.