OBS网络中IP分组长度分布对业务流自相似性的影响

近年来，大量研究发现，网络流量呈现自相似性，自相似会导致数据分组延迟增大，丢包率增加，网络性能 恶化。着重论述了在OBS固定时间汇聚算法下，用于汇聚成突发数据分组的IP分组长度对业务流自相似影响的 问题。突发数据分组的汇聚算法是光突发交换的关键技术之一，其中，固定时间汇聚（FAP）算法可以平滑流量、降 低流量自相似性。用于汇聚成突发数据分组的IP分组，其自身长度对固定汇聚（FAP）算法的流量整形有一定影 响。仿真结果表明，存在最佳汇聚时间门限，能极大地降低业务流自相似程度，且最佳汇聚时间门限与IP分组长

Impact of IP packet length distribution on self-similarity of traffic in OBS networks

The recent studies show that network traffic presents self-similarity. The self-similarity brings down the network efficiency and performance, increasing the data packet delay and the packet dropping probability. The impact of IP packet length assembled into a burst on self-similarity of traffic is elaborated under fixed-assembly-period (FAP) algorithm of optical burst switching (OBS). Burst assembly algorithms are one of the critical techniques of OBS. Among these algorithms, the fixed-assembly-period algorithm can smooth traffic and degrade traffic self-similarity. The simulation shows that the average length of IP packets assembled into a burst has impact on the traffic shaping of the FAP algorithm to some extent. The simulation result demonstrates the existence of the optimum fixed assembly period which can reduce the degree of self-similarity greatly. Furthermore, it reveals that there is a close relationship between the average length of IP packets and the optimum fixed assembly period.