一种基于Trie的流水式IP查找结构  

A Pipelined IP Lookup Structure Based on Trie

作　　者：张建伟[1] 滕飞[1] 马万里[1] 陈晓明[1] 

机构地区：[1]大连理工大学电子科学与技术学院,辽宁大连116024

出　　处：《微电子学与计算机》2017年第7期70-74,共5页Microelectronics & Computer

基　　金：国家自然科学基金项目(61306091;61340050);中央高校基本科研业务费专项资金(DUT15QT46)

摘　　要：为了减小基于Trie的流水式IP查找引擎的内存占用和延时,基于优化线性流水线(OLP),利用其引入空操作(Nop)能够更灵活布局的特点,提出了EOLP(Enhanced Optimized Linear Pipeline)方案,包括前缀扩展长度确立算法,流水线长度确立方案,布局算法和编码算法.实验结果表明,相对于OLP方案,EOLP方案的内存占用和流水线延时至少减少了1/6,能够完全布局Trie节点,并达到了72Gb/s的吞吐率,满足高速网络的需求.In order to reduce the memory usage and delay of pipelined IP lookup structure based on Trie, this paper presents a new solution called EOLP（Enhanced Optimized Linear Pipeline）, including prefix expansion algorithm, pipeline conforming algorithm, mapping algorithm and encoding algorithm, based on OLP that implements the NOP in the pipeline. Experimental results shows the memory usage and delay can be decreased by 1/6 and all the nodes can be mapped on. It achieves a throughput of 72Gbs, meeting the requirement of high-speed networks.

关 键 词：IP地址查找 最长前缀匹配 FPGA Trie流水线 

分 类 号：TN791[电子电信—电路与系统]