基于长相关的网络节点有效性研究

Research on availability of network node based on long-range dependence

针对因节点失效而产生的网络拥塞现象,提出了一种度量节点有效性的评价指标。该指标针对实际流量的长相关特性,分别利用自回归滑动平均(auto-regressive and moving average,ARMA)模型和线性分形稳定噪声(linear fractionalstable noise,LFSN)模型建立到达流量的预测方法。深入研究了该指标与平均到达速率之间的关系,仿真实验结果表明,当平均到达速率较小时ARMA模型性能较好,反之则LFSN模型性能较好。

In order to mitigate the network congestion by node failures,a novel availability evaluation indicator is proposed.In this indicator,based on the long-range dependence of actual traffic,the prediction methods of arrival traffic are established with auto-regressive and moving average（ARMA） model and linear fractional stable noise（LFSN） model respectively.Then,the simulation is conducted to research on the relationship between proposed indicator and average arrival rate.The result shows that the performance of ARMA model is better when the average arrival rate is small,and vice versa the performance of LFSN model is better.