一种高效的光分组交换网络缓存结构

An Effective Buffering Architecture for Optical Packet Switching Networks

提出了一种新型的、用于光分组交换网络的光纤缓存结构。该结构针对光分组交换网络中解决输出端口竞争需要大量光纤延迟线,从而导致核心节点造价太高的缺陷,采用光纤分段式共享的机制,充分提高光纤延迟线的利用率。仿真和分析表明这种新型结构能有效降低丢包率,并可以取得令人满意的光分组平均延时参数;随着光纤延迟矩阵规模的增大,该结构仅需增加很少数目的光纤延迟线单元便可实现核心节点的扩展;在使用的总光纤延迟线单元数目上,该结构更能体现其优越性。

A novel output buffer architecture for optical packet switching networks is proposed. The architecture which adopts fiber sharing mechanism and improves utilization of fiber delay lines, aims at solving the problem of requiring a large number of fiber delay lines which are used to solve resource contention but result in costliness in core node in optical packet switching networks. Simulation results and analysis show that the architecture can decrease packet loss probability effectively and achieve reasonable average delay per packet; when fiber delay matrix grows, the number of fiber delay lines will not grow obviously, this is favorable to extend of core node. At last, superiority of this architecture will be shown by comparing with simple output buffer architecture.