【作者】 魏斌；

【导师】 吴重庆；

【作者基本信息】 北京交通大学 ， 通信与信息系统， 2008， 博士

【摘要】 随着数据业务的不断增长,分组交换的需求越来越大,尤其是全光网络中的分组交换对于克服光电转换瓶颈、提高交换容量、延长分组的直达距离有着重要的意义。IP做为越来越多业务的承载方式已经应用于众多的网络中。与此同时由于小粒度异步分组交换所带来的业务的自相似性(突发性)也随着网络容量的升级和业务的不断拓展而愈加明显,并成为网络性能恶化的主要因素之一。本文针对异步全光分组交换网络中若干关键技术进行研究,主要内容如下:1.在深入分析自相似业务流的基础上,提出了一种快速Hurst指数估值过程,首次利用经验模态分解去除了影响Hurst指数估值的趋势项,并利用小波分解对于待估值序列进行了预处理,不但提高了估值精度,还大大加快了估值速度。同时基于当前网络研究中缺乏实用高速自相似流源的现状,在详细分析了各种自相似模型的特点和硬件平台特点的基础上,提出并研制成功了一种实时在线产生自相似业务的发生源。同时提出了基于数字逻辑的用于计算重尾分布随机数的单精度浮点数的幂次运算等相关算法。2.通过对数据封装算法的研究和分析,提出用于全光分组交换OPS的QHTAA算法来实现数据分组长度和分组发送时间间隔两个层面的平滑。数据分组长度的剧烈变化通过QHTAA算法的分组组装的门限来控制。而分组发送时间间隔则通过QHTAA算法引入的延迟因子DE来调整。通过这两个层面的调整使得分组单位时间内的流量的自相似性被大大的平滑掉了,表明QHTAA算法具有很好的抑制突发性的效果。3.针对全光边缘节点的接入,提出了一种新型的基于FDLs的边缘节点模型,同时针对该节点模型提出了一种新的主动队列维护算法,在一定程度上节省了FDLs缓存资源,并在统计FDLs信道端口使用情况的基础上提出了一种新的FDLs设计方案——有选择优化方案。该设计方案具有较低的丢包率和较小的数据包平均延迟,同时降低了边缘汇聚节点的硬件复杂度和资源消耗,为FDLs的设计提供了一种新的思路。4.针对试验室研制的含有DLOBs全光缓存器的全光分组交换节点提出了一种实现全光分组交换冲突解决的硬件调度设计方案和调度算法,并对DLOBs的最大环长(缓存深度)、最大缓存圈数和DLOB的个数(缓存单元端口数)进行了深入分析,得到:DLOBs器件的研究方向为缓存圈数(缓存时间)的增加,而深度根据QoS的要求来确定,DLOBs的个数不要超过10个。5.定量分析了光弹性分组环网(O-RPR)内的数据流量。并在对一种新的故障场景深入分析的基础上,提出一种可控源路由保护机制,弥补了原协议中动态保护机制的不足,提高了O-RPR的生存性。更多还原

【Abstract】 With the increasing of the data traffic, the need of packets switching is also increased fast. Especially in the optical packet switching network and the research on that field is becoming hot. More and more services transmitted over IP in all kinds of netwoks. At the same time, the self-similar of the data traffic, which caused by small granularity, asynchronous packets switching and the updating of network capacity is obvious. This thesis investigates on some key technologies of asynchronous optical packet switching network.1. Base on the investigation of the self-similar data traffic, a fast evaluating process of Hurst parameter has been put forward. It is the first time to get rid of the trend term and use the wavelet preprocesses data sequence. The process not only improves the evaluation precision but also decrease the time consuming. According the state of lacking of self-similar generator for the core node performance testing, a real self-similar data traffic source has been developed and algorithms for Pareto distribution random value base on signal float number and power calculation have been put forward.2. Put forward a QHTAA algorithm to realize the traffic smoothing by controlling both packet length and packet output interval. The burst of packet length will be controlled by packet encapsulation threshold. And the packet output interval will be controlled by delay factor DE. After the packet encapsulation of the edge node the burst of the data traffic has been restrained well.3. Put forward a new opical edge node structure for optical packet access. And an algorithm base on active queue management has been put forward to improve the efficiency of the fiber channel. Also a scheme of FDLs arrangement which named chosen optimization arrangement proved its good performance of lower packet loss rate, small average packet delay and simplifies the complexity of the hardware and resource consuming.4. Put forward a schedule algorithm and hardware design to realize the optical packet switching base on the DLOB buffer which developed by our lab. And base on the investigation of the switching node, the maximum loop should design according to the need of QoS and the number of the DLOBs should less than 10.5. A model of optical resilient packet ring has been built and it is used to investigate the data traffic which transmitted over the O-RPR network. And under a new failure scenario, the protect mechanism has been put forward as a complement of the original protocol.更多还原